40 Years of Forrest and Tomlin

40 Years of Forrest and Tomlin

Tuesday, 8:00am - 9:30am

TA01

01- West 101- CC

Joint Session Optimization Global/Integer: Piecewise

Linear Functions in Global Optimization and Mixed

Integer Programming

Sponsor: Optimization/Global Optimization & Optimization/Integer

Programming

Sponsored Session

Chair: Juan Pablo Vielma, University of Pittsburgh,

1043 Benedum Hall, 3700 O’Hare Street, Pittsburgh, PA, 15261,

United States of America, [email protected]

1 - A Global Optimization Framework for Mixed-Integer

Signomial Programs

Christodoulos Floudas, Stephen C. Macaleer ‘63 Professor in

Engineering and Applied Science, Professor of Chemical and

Biological Engineering, Princeton University, Department of

Chemical & Biological Eng., Princeton, NJ, 08544,

United States of America, [email protected], Ruth Misener

Expanding the Global Mixed-Integer Quadratic Optimizer, GloMIQO, we develop a framework for addressing mixed-integer signomial optimization problems to epsilon-global optimality. Our approach reformulates user input, detects special mathematical structure, and globally optimizes the transformed problem. This presentation highlights dynamically generated cutting planes based on convexification strategies including alphaBB, RLT, and specialized underestimators for specific functional forms.

2 - Towards Globally Optimal Solutions for MINLPs by

Discretization Techniques

Björn Geifller, Dr., FAU Erlangen-Nürnberg, Cauerstr. 11,

Erlangen, 91058, Germany, [email protected],

Alexander Martin, Antonio Morsi, Lars Schewe

We show how to turn any general purpose MIP-solver into a general purpose solver for MINLPs. To this end we construct arbitrary tight MIP-relaxations of the underlying nonlinear problem by extending some well-known MIP-techniques for piecewise linear functions. We present an algorithm which is based on an adaptive refinement of these relaxations and provide numerical evidence for the success of our approach by means of problems from gas network optimization.

3 - Solving Non-linear Optimization Problems using Piecewise

Linear Approximation

Vivek Vasudeva, Student, University of Texas at Austin,

Department of Information, Risk and Op. Mgmt., 1 University

Station, B6000, Austin, TX, 78712, United States of America, [email protected], Anant Balakrishnan,

Leon Lasdon

This talk discusses a method that combines mixed-integer programming and local search to solve non-linear optimization problems. The approach entails first solving a piecewise linear approximation to the original problem in order to obtain a solution that lies in the basin of attraction of the global solution, and then using this solution as a starting point for a local solver. We discuss variants and enhancements of this approach and report results for selected non-convex non-linear problems.

4 - Incremental Formulations for SOS1 Variables

Sercan Yildiz, Graduate Research Assistant, University of

Pittsburgh, 1048 Benedum Hall, 3700 O’Hara Street, Pittsburgh,

PA, 15261, United States of America, [email protected],

Juan Pablo Vielma

We introduce a hierarchy of formulations for SOS1 variables that incorporates the incremental and logarithmic formulations as its extreme cases. We use these formulations to build MILP models for piecewise linear functions and present computational results that allow us to compare their relative effectiveness.

INFORMS Phoenix – 2012

TA03

TA02

02- West 102 A- CC

Economics of Sequential Decision Making

Sponsor: Decision Analysis

Sponsored Session

Chair: H. Dharma Kwon, University of Illinois at Urbana-Champaign,

1206 South Sixth Street, Champaign, IL, 61820,

United States of America, [email protected]

1 - Strategic Investment under Spillovers and

Information Externalities

Wenxin Xu, University of Illinois, 350 Wohler’s Hall, Champaign,

IL, 61820, United States of America, [email protected], H.

Dharma Kwon, Anupam Agrawal, Suresh Muthulingam

We study the problem of two firms with investment opportunities when the true economic value of the return is unknown. In particular, we explore the impact of the spillover effects of investment and the information externalities, which arise when the one who invests later can observe the leader’s profit stream and learn about the true economic value of the investment opportunities. We find that their impact on the timing of investment is non-trivial.

2 - Learning Consumer Tastes: A Nonparametric Bayesian Model

Canan Ulu, Assistant Professor, University of Texas at Austin,

2110 Speedway Stop B6500, Austin, TX, United States of America,

[email protected], Dorothee Honhon

We develop a nonparametric Bayesian learning model for a firm that can gather information about consumer tastes through sales of its product assortment. The firm can dynamically change its product assortment from period to period to gather better information about consumer tastes. We consider heuristics to solve large problems.

3 - Competitive Control of Market Goodwill and Strategic Exit

Hongzhong Zhang, Assistant Professor, Columbia University, New

York, NY, 10027, United States of America, [email protected],

H. Dharma Kwon

We consider the problem of two firms competing for market goodwill, which randomly evolves in time. Each firm can make irreversible investment to increase its goodwill to improve its profit stream and decrease its rival firm’s goodwill.

They also have an option to exit the market at any point in time. We formulate the problem as a stochastic singular control and stopping game, and we obtain a rich variety of Markov perfect equilibria.

4 - Using Nash Bargaining to Design Project Management

Contracts under Cost Uncertainty

Steven Lippman, Professor, University of California-Los Angeles,

Anderson School, 110 Westwood Plaza, Suite D517,

Los Angeles, CA, 90095-1481, United States of America, [email protected], Kevin McCardle,

Christopher Tang

For procurement contracts in which the cost is uncertain, cost sharing is common.

We determine the best cost-sharing contract between a risk-neutral project manager and a risk-averse contractor when negotiation proceeds in accord with

Nash bargaining.

TA03

03- West 102 B- CC

Joint Session DAS/MAS: Game Theory and Homeland

Security I - Screening Models

Sponsor: Decision Analysis & Military Applications

Sponsored Session

Chair: Jun Zhuang, Assistant Professor, University at Buffalo, SUNY,

403 Bell Hall, Buffalo, NY, 14260, United States of America, [email protected]

1 - PROTECT: A Deployed Game Theoretic System to Protect the

Ports of the United States

Milind Tambe, University of Southern California, Los Angeles, CA,

United States of America, [email protected], Eric Shieh,

Bo An, Rong Yang, Fei Fang, Albert Xin Jiang, Craig Baldwin,

Joseph DiRenzo, Ben Maule, Garrett Meyer

PROTECT is a game-theoretic system deployed by the United States Coast Guard in the port of Boston for scheduling their patrols. PROTECT offers several key innovations. First, this system is a departure from the assumption of perfect adversary rationality, relying instead on a quantal response (QR) model of the adversary. Second, We present two new algorithms for computing defender strategies in this QR model. Third, we generate a compact representation of the defender’s strategy space.

261

TA04

2 - Multiple-stage Screening Game in the Face of

Strategic Applicants

Cen Song, Teaching Assistant, University at Buffalo, 435 Bell Hall,

Buffalo, NY, United States of America, [email protected],

Jun Zhuang

The security screening system becomes a big concern to identify and deter potential attacks. Based on analyzing the security and congestion in the face of strategic applicants, this paper is further studied in an imperfect and multi-stage screening game. In series queueing network, the error screen probabilities are considered. We provide analytical optimal level of screening strategies for the approver and compare the discriminatory and non-discriminatory screening policies.

3 - Robust Screening Considering the Applicant’s Behavior

Jie Xu, University at Buffalo, 435 Bell Hall, Buffalo, NY, 14260,

United States of America, [email protected], Jun Zhuang

The performance of an optimal screening strategy depends critically on how accurately the applicants’ behavior has been predicted. In general, players’ behavior may be irrational, non-strategic, misinformed or biased. Thus, we use robust optimization to study screening policy with uncertainty about the applicants’ strategic ability, utility functions, and constraints, and how sensitive their optimal decisions and expected payoffs are to those uncertainties and the desired level of robustness.

4 - US Visa Applicant Wait Time Analysis

Marie Catalano, Undergraduate Student, University At Buffalo,

435 Bell Hall, Buffalo, NY, 14260, United States of America, [email protected], Elizabeth Newell, John Coles, Jun Zhuang

The wait time for US Visa applicants to receive a visa varies dramatically often causing excessive and unnecessary wait times. This research analyzes over 5,000 data points and observes the greatest number of people apply in June and

December (from 2008-2012). Trends shown by varying visa type, major, month and year in this research also provide insight to the Visa process.

TA04

04- West 102 C- CC

Data Envelopment Analysis I

Cluster: Data Envelopment Analysis

Invited Session

Chair: Timo Kuosmanen, Professor, Aalto University School of

Economics, PL 21220, Helsinki, Finland, [email protected]

1 - Market Demand Analysis by using the Market-data

Envelopment Analysis Model, The Case Study: Analysis

Mehdi Ahmadpanah, [email protected], Maryam Ghanizadeh

This paper introduces a market demand analysis model using the Market-Data

Envelopment Analysis technique. M-DEA is a new approach that initially uses the

AHP technique to extract the main and affecting parameters on different regions’ demand and then according to the extracted parameters, it assesses market demand in each region. Among the characteristics of this model is that the effective parameters are considered as a combination to evaluate the market volume of different regions.

2 - Estimating a Nonparametric Non Input Homothetic S-shaped

Production Function

Ole Olesen, Professor, University of Southern Denmark,

Campusvej 55, Odense M, DK-5230, Denmark, [email protected],

John Ruggiero

The use of the convex hull estimation in data envelopment analysis (DEA) models is often maintained, although standard microeconomic production theory posits a nonconvex S-shaped production frontier. Recently for an input homothetic production function a nonparametric estimation approach that allows for a nonconvex S-shaped scaling law has been proposed. In this paper we generalize this approach to allow for certain non input homothetic isoquants.

3 - What is the Best Practice for Yardstick Regulation of

Electricity Distribution?

Timo Kuosmanen, Professor, Aalto University School of

Economics, PL 21220, Helsinki, Finland, [email protected],

Antti Saastamoinen, Timo Sipilainen

Electricity distribution is a natural local monopoly. Most regulators apply data envelopment analysis (DEA) or stochastic frontier analysis (SFA) to estimate efficient cost. In Finland, the regulator adopted StoNED method in 2012. This paper compares DEA, SFA and StoNED using data of electricity distribution.

While efficiency scores are highly correlated, the cost targets show major differences. StoNED yields superior precision in Monte Carlo simulations.

INFORMS Phoenix – 2012

TA05

05- West 103 A- CC

Functional and Profile Data Analysis; Methods and

Applications

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Kamran Paynabar, University of Michigan, Ann Arbor, MI,

United States of America, [email protected]

1 - A Probabilistic Prediction Framework for Personalized Online

Prediction of Epileptic Seizures

Shouyi Wang, University of Washington, 3311 NE 65TH Street,

Seattle, WA, 98115, United States of America, [email protected],

Art Chaovalitwongse

The talk presents a novel probabilistic framework which integrates and innovates data mining, machine learning and probability theory for online pattern recognition and prediction of target events from nonstationary time series data.

With an attractive adaptive learning capability, the proposed probabilistic framework has been demonstrated to be effective to achieve a personalized online seizure prediction for patients with epilepsy.

2 - On using Profile Monitoring Techniques for Monitoring Point

Cloud Data

Fadel Megahed, Assistant Professor, Auburn University,

3301L Shelby Center, Auburn, Al, United States of America, [email protected]

3D laser scanners can rapidly provide point clouds consisting of millions of data points, which depict the entire surface geometry of a manufactured part.

Consequently, 3D scan data have a great potential to detect unexpected faults.

However, SPC methods capable of handling these large data-sets are yet to be developed. In this talk, we show how 3D scanners can be used to significantly improve the monitoring capabilities for manufacturing parts, characterized by complex surface geometries.

3 - Statistical Process Screening System: An Approach For

Identifying Irregular Individuals

Peihua Qiu, Professor, University of Minnesota,

224 Church Street SE, 313 Ford Hall, Minneapolis, MN, 55455,

United States of America, [email protected], Dongdong Xiang

We often need to identify individuals whose longitudinal pattern in certain indices is different from the pattern of those normally working individuals. In many applications, observations of a given individual are obtained sequentially, and it is desirable to have a screening system to give a signal of irregular longitudinal pattern as soon as possible. This paper proposes a method for that purpose.

TA06

06- West 103 B- CC

Simulation and Optimization

Contributed Session

Chair: Berna Dengiz, Baskent University, Engineering Faculty Baglica

Campus, Eskisehir Road 20th km, Ankara, Turkey, [email protected]

1 - A Dynamic Network Oligopoly Model with Product

Differentiation and Quality Competition

Dong Li, Univerity of Massachusetts-Amherst, 121 Presidents

Drive, Amherst, MA, 01003, United States of America, [email protected], Anna Nagurney

In this paper we develop a dynamic model of oligopolistic competition that includes product differentiation and quality levels in a network framework. We also provide stability analysis results and an algorithm with convergence results, along with numerical examples. This framework can serve as the foundation for the modeling and analysis of competition among firms in industries ranging from food to high tech products.

2 - Analysis of the Effects of Flexibility Factors on Supply

Chain Performance

Seratun Jannat, Graduate Student, Mississippi State University,

Department of Industrial and System Eng., P.O. Box 9542,

Starkville, MS, 39762, United States of America, [email protected], Allen Greenwood

In the design of stochastic flexible supply chains it is important to consider the effects on lead time and cost of such factors as the level of supplier flexibility, efficiency of alternative suppliers, information systems automation, and priority levels. This paper uses discrete-event simulation to assess the impact of these factors and recommend combinations of factor values that positively effect lead time and cost.

262

3 - Improving the Processing of Clinical Trial Agreements with

Simulation and Optimization

Han Wu, Student, University of Louisville, Department of

Industrial Engineering, JB Speed School of Engineering,

U of L, Louisville, KY, 40292, United States of America, [email protected], Prajwal Khadgi, Gerald W. Evans,

Sunderesh Heragu

The processing of clinical trial agreements (e.g., for the testing of new drugs) at universities is complex, involving the drug company, hospital(s), and university review boards. Reducing the time required for this process is important, since typically there will be competition among various institutions for the agreement with the drug company. This presentation will describe how a simulation model was developed to streamline the process at the University of Louisville.

4 - A TOPSIS- Based Taguchi Approach for Multi Response

Simulation Optimization

Berna Dengiz, Baskent University, Engineering Faculty Baglica

Campus, Eskisehir Road 20th km, Ankara, Turkey, [email protected], Yusuf Tansel Ic

In this study, a multi-response simulation optimization problem related the flexible manufacturing system (FMS) is solved using the TOPSIS based Taguchi method. In the solution procedure, this method reduces the multi-response optimization problem to a single-objective decision-making problem, because

Taguchi method is not able to handle this kind of simulation optimization problems only by itself. The hybrid simulation optimization method is applied to a real problem.

TA07

07- West 104 A- CC

INFORMS Phoenix – 2012

Joint Session DM/QSR: Bayesian Inference and

Computational Issues

Sponsor: Data Mining & Quality, Statistics and Reliability

Sponsored Session

Chair: Tevfik Aktekin, Assistant Professor of Decision Sciences,

University of New Hampshire, 15 Academic Way, Durham, NH, 03824,

United States of America, [email protected]

1 - Tuberculosis and HIV Mortality, Modelling of Multivariate ZIP

Distribution with Time Series

Rasim Musal, Assistant Professor, Texas State University, 1611

West Fifth Street, Austin, TX, 78703, United States of America, [email protected], Tevfik Aktekin

We investigate the effect of poverty and inequality on HIV and Tuberculosis mortality in the 62 counties of NY with spatial effects. We quantify inequality with the Theil Index and poverty via the ratios of the two Census variables in each ZCTA. As with HIV, Tuberculosis is a rare cause of mortality and many of the counties have 0 counts. This provides the motivation for a joint zero inflated

Poisson distribution Bayesian model with time.

2 - Applications of Bayesian Methods in Healthcare Fraud

Detection

Tahir Ekin, PhD Candidate, The George Washington University

Department of Decision Sciences, 2201 G St Funger Hall 415 NW,

Washington, DC, 20052, United States of America, [email protected],

Toros Caglar, Refik Soyer

Some applications of Bayesian ideas in healthcare fraud detection will be presented. The emphasis will be on the use of Bayesian co-clustering and link analysis methodologies to identify billing patterns of potentially fraudulent providers who submit a large number of high-dollar claims submitted to healthcare insurance programs for reimbursement. Alternative Bayesian frameworks for modelling sample-based audits of healthcare related payments and overpayment situations will also be discussed.

3 - Dynamic Bayes Network Approach to Market Basket Analysis

Bumsoo Kim, PhD Candidate, George Washington University,

2201 G Street NW, Funger Hall #415H, Washington, DC, 20052,

United States of America, [email protected], Srinivas Prasad,

Refik Soyer

In this paper we consider utilizing DBN’s to model different market basket problems. The proposed approach can be used for learning and inference objectives, for both model parameters of a given structure and the dynamics of the structure itself. We use MCMC methods and use results from the DBN literature to devise learning algorithms in a time-series market basket data setting.

We illustrate the implementation of proposed approach with low-dimensional problems and discuss their applications.

4 - Bayesian Parametric and Semi-parametric Modeling of Call

Center Abandonment

Tevfik Aktekin, Assistant Professor of Decision Sciences, University of New Hampshire, 15 Academic Way, Durham, NH, 03824,

United States of America, [email protected], Refik Soyer

We consider the modeling and inference of abandonment behavior in call centers.

We present different family of distributions, piecewise time to abandonment, mixtures and semi-parametric models, develop Bayesian inference for posterior and predictive analyses with censored abandonment and discuss implications on staffing. We use real call center data, present additional insights that can be obtained from the Bayesian analysis and discuss implications for different customer profiles.

TA08

08- West 104 B- CC

Emergency Medical Services

Sponsor: Public Programs, Service and Needs

Sponsored Session

TA08

Chair: Laura McLay, Virginia Commonwealth University, Statistics &

Operations Research, 1015 Floyd Ave, Box 843083, Richmond, VA,

23284, United States of America, [email protected]

1 - Next Generation Emergency Location Tools: Heuristics with

Simulation

Hari Rajagopalan, Associate Professor, Francis Marion University,

P.O. Box 100547, Florence, SC, 29502, United States of America,

[email protected], Cem Saydam, Elizabeth Sharer,

Muhammad Zaffar

Location models for emergency response systems have rapidly evolved from the initial mathematical models which required many simplifying assumptions to make the models tractable. In this paper, we discuss the use of simulation with meta-heurstic search methods in location models to bring more realism and generate a richer output to get a better insight on the effectiveness of location models.

2 - EMS Planning and Management

Armann Ingolfsson, University of Alberta, School of Business,

Edmonton, AB, T6G2R6, Canada, [email protected]

I survey research on planning and management for Emergency Medical Services, including: Forecasting demand, forecasting response times, measuring performance, choosing base locations, allocating ambulances to bases, based on predictable and unpredictable changes in demand and travel times. I focus on empirical work, the use of analytical stochastic models, and ways in which medical researchers and operations researchers could benefit from collaboration.

3 - Joint Location and Dispatching Decisions for Emergency

Medical Services

Hector Toro-Diaz, Clemson University, 204 Freeman Hall,

Clemson, SC, 29634, United States of America, [email protected], Maria Mayorga, Sunarin Chanta,

Laura McLay

Reducing ambulance response time saves lives. We propose a joint location and dispatching model aimed to minimize the Mean System Response Time (MRT), and a solution procedure based on Genetic Algorithms (GA). The commonly used fixed a-priori dispatching rule sending the closest ambulance has shown to be effective in minimizing the MRT across a variety of different system’s characteristics.

4 - Predictive Performance of Data Mining and Regression

Methodologies on 911 Call Counts

Laura McLay, Virginia Commonwealth University, Statistics &

Operations Research, 1015 Floyd Ave., Box 843083, Richmond,

VA, 23284, United States of America, [email protected],

Richard Garrett

In this presentation, we compare data-mining and regression models for forecasting emergency medical and fire 911 call counts. A novel aspect of this project is the inclusion of weather variables in the data set. The data mining techniques include Support Vector Machines (SVM) using the Gaussian kernel and Random Forests (RF). The data mining techniques are compared to negative binomial regression. The predictive performance of all models is assessed via 10fold cross-validation.

263

TA09

TA09

09- West 105 A- CC

Advances in Multiobjective Programming

Sponsor: Multiple Criteria Decision Making

Sponsored Session

Chair: Margaret Wiecek, Professor, Clemson University, Martin Hall

O-208, Clemson, SC, 29634, United States of America, [email protected]

1 - Block Coordinate Descent-type Decomposition in

Multiobjective Programming

Brian Dandurand, Clemson University, O-110 Martin Hall,

Box 340975, Clemson, SC, 29631, United States of America, [email protected], Margaret Wiecek

Decomposition of multiobjective programs (MOPs) in the objective space, the decision space, and simultaneously both spaces according to the principles of block coordinate descent is investigated. Coordination algorithms and conditions for their convergence to the Pareto set of the original MOP are developed.

Computational examples are given.

2 - Robust Multiobjective Optimization: Theory and Methodology

Erin Doolittle, Clemson University, Department of Mathematical

Sciences, Clemson, SC, 29634, United States of America, [email protected], Margaret Wiecek, Hervé Kerivin

For an uncertain multiobjective optimization problem (MOP) we develop conditions for the feasibility of a related robust counterpart. We show that MOPs with uncertainty in the objective functions or the right-hand side of the constraints are equivalent to MOPs with uncertainty in the constraint coefficients.

For data uncertainty represented as ellipsoidal sets, we transform the uncertain

MOP into a tractable and deterministic MOP, keeping the same Pareto solution set.

3 - Multiplicative Epsilon-Dominance in

Multiobjective Programming

Lakmali Weerasena, PhD Student, Clemson University, O-110,

Martin Hall, Box 340975, Clemson, SC, 29634, United States of

America, [email protected], Margaret Wiecek

Multiplicative epsilon-dominance is studied as a binary relation that yields the epsilon-Pareto set being an approximation of the Pareto set for multiobjective programs (MOPs). The effect of generalizing the Pareto cone to polyhedral or more general convex cones in order to obtain approximations of nondominated sets of MOPs is examined.

4 - Advancing Branch-and-bound Techniques for Mixed-integer

Multiobjective Programming

Margaret Wiecek, Professor, Clemson University, Martin Hall

O-208, Clemson, SC, 29634, United States of America, [email protected], Pietro Belotti, Banu Soylu

Properties of the Pareto set for mixed integer linear multiobjective programs

(MOPs) are studied. Branch-and-bound (BB) rules for pure integer MOPs are reviewed and examined against their applicability for mixed integer MOPs. More effective components of a BB algorithm, namely a fathoming procedure and a data structure for Pareto solutions, especially tailored to mixed integer MOPs, are proposed and integrated into a BB algorithm.

TA10

10- West 105 B- CC

Stochastic Financial Optimization

Sponsor: Optimization/Stochastic Programming

Sponsored Session

INFORMS Phoenix – 2012

Chair: Miguel Lejeune, George Washington University, 2201 G Street,

NW, Washington, DC, 20052, United States of America, [email protected]

1 - Library of Test Financial Optimization Problems at the

University of Florida

Stan Uryasev, American Optimal Decisions, 5214 SW 91 Way, Ste.

#130, Gainesville, FL, 32608, United States of America, [email protected]

Presentation will describe the library of stochastic financial optimization problems posted at the University of Florida at this link: www.ise.ufl.edu/uryasev/research/testproblems/financial_engineering/ The library includes problem statements, data, and solutions in Portfolio Safeguard format.

2 - Percentile Optimization in Portfolio Management

Aurelie Thiele, Associate Professor, Lehigh University, [email protected], Elcin Cetinkaya

We investigate portfolio management problems where the decision-maker seeks to maximize a left-tail percentile of the return distribution. We discuss datadriven techniques to incorporate the percentile in a tractable formulation and present insights into the resulting strategy, both from a theoretical standpoint and in numerical experiments.

3 - Risk-Averse Enhanced Indexation

Miguel Lejeune, George Washington University, 2201 G Street,

NW, Washington, DC, 20052, United States of America, [email protected], Gulay Samatli-Pac

We propose a partial replication strategy to construct risk-averse enhanced index funds. The variance of the index fund return must be below a low-risk threshold with a large probability, thereby limiting the market risk exposure. We derive a deterministic equivalent for the quadratic risk constraint and develop an exact outer approximation method. The method provides a hierarchical organization of the computations with expanding sets of integer-restricted variables and scales well.

TA11

11- West 105 C- CC

First-Order Methods for Large-scale Convex

Optimization and Their Applications

Sponsor: Optimization/Nonlinear Programming

Sponsored Session

Chair: Fatma Kilinc-Karzan, Assistant Professor, Carnegie Mellon

University, 5000 Forbes Avenue, Pittsburgh, PA, 15213,

United States of America, [email protected]

1 - On Unified View of Nullspace-type Conditions for Sparse and

Low-rank Recoveries

Fatma Kilinc-Karzan, Assistant Professor, Carnegie Mellon

University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United

States of America, [email protected], Arkadi Nemirovski,

Anatoli Juditsky

We present general framework to handle sparsity structures and associated recoveries of signals from their linear image of reduced dimension possibly corrupted with noise. This unified treatment includes sparse and block-sparse recoveries as well as low-rank matrix reconstruction. We present sufficient conditions for the recovery to be precise in the noiseless case, derive error bounds for imperfect recovery under these conditions and present efficiently verifiable sufficient conditions.

2 - Accelerating Level Methods for Large-scale

Convex Optimization

Guanghui Lan, University of Florida, 303 Weil Hall, Gainesville,

FL, 32611, United States of America, [email protected], Cong Dang

We introduce accelerated level methods optimal for large-scale convex programming problems and smoothing level methods optimal for a class of structured saddle point problems. Promising numerical results for solving certain

Semidefinite programming and imaging problems will be presented.

3 - Accelerated Proximal-Gradient Homotopy Method for the

Sparse Least-Squares Problem

Qihang Lin, PhD Student, Tepper School of Business, Carnegie

Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213,

United States of America, [email protected], Lin Xiao

We solve L1-regularized least-squares (L1LS) problem by approximately solving a sequence of L1LS with decreasing regularization parameters. Each L1LS is solved by accelerated gradient method and the returned solution is used to warm start the next L1LS. Under suitable assumptions, this method ensures that all iterates are sparse so that it obtains a global geometric convergence rate due to the strongly convexity of the objective function restricted on the sparse subspace.

264

INFORMS Phoenix – 2012

TA12

12- West 106 A- CC

Theory and Application of Mixed Integer

Programming

Sponsor: Optimization/Integer Programming

Sponsored Session

Chair: Kiavash Kianfar, Assistant Professor, Texas A&M University,

TAMU 3131, College Station, TX, 77843-3131, United States of

America, [email protected]

1 - Network Design with Stochastic Arc Capacities

Avinash Bhardwaj, University of California Berkeley, 450/60

Sutardja Dai Hall, University of California Berkeley, Berkeley, CA,

94720-1764, United States of America, [email protected], Alper Atamturk

We present models and branch-and-cut algorithms to find a min cost fixed charge network with stochastic arc capacities. We derive strong valid inequalities for the convex hull of the solutions. We provide inequality extension and lifting algorithms to obtain the facets of the convex hull.

2 - An Integrated Approach for the Fleet Assignment, Aircraft

Routing, and Crew Pairing Problem

Shengzhi Shao, Virginia Tech, 250 Durham Hall (0118),

Blacksburg, VA, 24061, United States of America, [email protected],

Hanif D. Sherali, Mohamed Haouari

We propose a novel formulation that integrates fleet assignment, aircraft routing, and crew pairing within a single mixed-integer linear program, and derive valid inequalities to further tighten the representation. A Benders decomposition-based solution approach is adopted together with several acceleration strategies.

Computational results are presented using real-life data obtained from a major

U.S. airline to demonstrate the efficacy of our approach and to provide insights.

3 - Polyhedral Studies for Vendor-Managed Inventory Models

Ayse Nur Arslan, PhD Student, University of Florida, Department of Industrial and Systems, Engineering 303 Weil Hall, Gainesville,

FL, 32608, United States of America, [email protected],

Jean-Philippe Richard, Yongpei Guan

We study a family of vendor-managed inventory models. We first present a polynomial time dynamic programming algorithm and the corresponding extended formulation. In addition, we perform a polyhedral study of the corresponding set. We obtain several families of facet-defining inequalities which provide the convex hull description for two periods.

4 - Continuing Work on Lifting Nonlinear Nonconvex Formulations

Daniel Bienstock, Columbia University, Department of IEOR,

New York, NY, United States of America, [email protected]

We describe continuing work on efficient cutting-plane procedures for addressing nonconvex feasible regions. Some computational results may be provided.

TA13

13- West 106 B- CC

Applications of Convex Optimization

Sponsor: Optimization/Linear Programming and Complementarity

Sponsored Session

Chair: Miguel F. Anjos, Canada Research Chair in Discrete Nonlinear

Optimization in Engineering, Ecole Polytechnique de Montreal, C.P.

6079, succ. Centre-ville, Montreal, QC, H3C 3A7, Canada, [email protected]

1 - QPCCs, QCQPs, and Copositive Programming

John Mitchell, Professor, Rensselaer Polytechnic Institute,

Math Sciences, Troy, NY, 12180, United States of America, [email protected], Jong-Shi Pang, Lijie Bai

We show that members of a certain class of quadratically constrained quadratic programs (QCQPs) can be represented as quadratic programs with complementarity constraints (QPCCs). We exploit this relationship to show that the Frank-Wolfe theorem holds for the members of this class of QCQPs, namely that they attain their optimal value if it exists. We also show that a completely positive relaxation of certain QPCCs is tight, even if the complementary variables in the QPCC are unbounded.

TA14

2 - Robust Low-rank Tensor Recovery via Convex Optimization

Zhiwei Qin, Columbia University, Room 313, 500 W. 120th St.,

New York, NY, 10027, United States of America, [email protected], Donald Goldfarb

Recovering low-rank tensors from gross corruptions and missing values is important for multilinear data analysis. We study the problem of robust low-rank tensor recovery in a convex optimization framework, and we propose efficient algorithms based on the alternating direction method of multipliers and the accelerated proximal gradient method. The practical effectiveness of the model and the algorithms is demonstrated through a series of synthetic and real examples.

3 - A Branch and Cut Algorithm for Solving Capacitated Max K-Cut with an Application in Scheduling

Matthew Oster, Rutgers University, 640 Bartholomew Road,

Piscataway, NJ, 08854, United States of America, [email protected], Jonathan Eckstein

We model the scheduling of a symmetric multi-track conference as a capacitated version of Maximum K-Cut (MKC). We solve this NP-hard problem to optimality within a branch-and-bound framework equipped with a semidefinite programming relaxation of MKC, enhanced with triangle and clique cuts, as well as problem-specific cuts (e.g. capacity cuts). We introduce a heuristic for generating feasible solutions at most tree nodes. Test results will be discussed for serial and parallel implementations.

4 - A Semidefinite Optimization Approach to Space-free Multi-row

Facility Layout

Miguel F. Anjos, Canada Research Chair in Discrete Nonlinear

Optimization in Engineering, Ecole Polytechnique de Montreal,

C.P. 6079, succ. Centre-ville, Montreal, QC, H3C 3A7, Canada, [email protected], Philipp Hungerlander

Multi-row facility layout seeks an optimal placement of departments along rows.

Large single-row problems have been solved to global optimality using semidefinite optimization. We extend the approach to space-free multi-row layout and show that if provides high-quality bounds in reasonable time.

TA14

14- West 106 C- CC

Planning and Scheduling Models for Manufacturing and Service Systems

Cluster: Scheduling and Project Management

Invited Session

Chair: Chelliah Sriskandarajah, University of Texas at Dallas, 800 W.

Campbell Rd. Sm30, Richardson, 75080, United States of America, [email protected]

1 - Project Management under Environmental Regulation

Gokce Esenduran, Ohio State University, College of Business,

656 Fisher Hall Fisher, Columbus, OH, 43210, United States of

America, [email protected], Nicholas Hall, Zhixin Liu

We investigate the implications of environmental regulations on project scheduling. We model the decisions of a regulator and multiple project managers

(PMs) as a Stackelberg game. The regulator specifies pollution remediation requirements, and the PMs respond either by remediation or by investment in greener task design. Our results show that the decisions of the regulator and PMs are not initially coordinated. However, the regulator can coordinate them by providing a subsidy to the PMs.

2 - A Special Hybrid Flowhsop Scheduling Problem with No Buffer

Esaignani Selvarajah, Assistant Professor, University of Windsor,

401 Sunset Avenue, Windsor, On, Canada, [email protected]

We study job sequencing in a special hybrid flowshop when there is no buffer.

The problem is NP-hard. Therefore, we first study the two stage hybrid flowshop problem. Then we provide a mathematical model and a heuristic algorithm for the multi-stage m-machine problem.

3 - A New Heuristic for Scheduling Strictly Periodic Services in

Discrete Time Intervals

Osman Kazan, Teaching Assitant, Jindal School of Management,

University of Texas at Dallas, 800 West Campbell Road,

Richardson, TX, 75080, United States of America, [email protected], Milind Dawande,

Chelliah Sriskandarajah, Kathryn Stecke

Operational practices of a national recycling and waste management company inspired us to formulate a combinatorial optimization problem. The challenge was to schedule strictly periodic services with a perfectly balanced total weekly workload. We analyzed the complexity and established heuristic techniques for well-balanced schedules with worst-case guarantees. The effectiveness of these techniques are confirmed by computational results on real-life and artificial data.

265

TA15

4 - Value of Information in a Three-Player Environment

Manoj Vanajakumari, Texas A&M University, 3367 TAMU,

Collge Station, TX, 77843, United States of America, [email protected], Subodha Kumar

In this research we study a supply chain system in which a manufacturer has two suppliers — a low cost supplier (S1) and a high cost supplier (S2). S1 has a dedicated capacity for the manufacturer; however, S2 allocates capacity for the manufacturer based on a forecast provided. In a given period, the manufacturer has to optimize the fraction of the demand allocated to S2 while S2 has to find the capacity allocated for the manufacturer. In such a setting we study the value of information.

TA15

15- West 202- CC

INFORMS Phoenix – 2012

Software Demonstration

Invited Session

1 - AnyLogic Simulation Software-Examples and Customer

Case Studies

George Gonzalez-Rivas, National Director, AnyLogic, 53 Frontage

Rd., Ste. 115, Hampton NJ 08827, United States of America, [email protected], Andrei Borshchev

Customer case studies and examples of AnyLogic. AnyLogic simulation software is compatible with Windows, MAC and Linux, allows you to build powerful simulation models and has 3D animation capabilities. AnyLogic works with three paradigms of modeling; system dynamics, agent based modeling and discrete event modeling.

2 - JMP Software from SAS - Visual

Mia Stephens, JMP Software from SAS, 100 SAS Campus Dr.,

Cary NC 27513, United States of America, [email protected]

JMP Statistical Discovery Software is interactive and visual desktop software for

Windows and Mac, with a complete array of integrated graphical and statistical procedures. We will demonstrate JMP tools for data analysis, visualization and exploration, including, Graph Builder®, bubble plots, the data filter, and our popular mapping tools.

This talk will recount the creation of the first spreadsheet program, VisiCalc, on the Apple II in 1979, including its development and marketing. At the time, experts believed that the need for financial planning and modeling software was fully served by several large competitors, with powerful systems offered through time-sharing. But the author and his colleagues saw the world differently. We now know that the experts were wrong 33 years ago - but what does that mean for today?

2 - Spreadsheet Shifting Sands: Is There an App for That?

Tom Groleau, Chair, Social Science Division, Carthage College,

2001 Alford Park Drive, Kenosha, WI, 53140,

United States of America, [email protected]

VisiCalc revolutionized end-user computing and created a generation of spreadsheet-native college students. As Excel took over the spreadsheet world, the OR/MS community used it as a bridge to bring modeling to the masses.

However, the landscape may be changing. Does “spreadsheet” still equal “Excel”?

As app-native students join the work world, will spreadsheets continue to be the dominant end-user modeling tool?

3 - What Do New Graduates Need to Know?

Rick Carter, CEO, Equation Consulting, 2650 S Decker Lake Blvd.,

Salt Lake City, UT, United States of America, [email protected]

Equation Consulting does spreadsheet based analysis for healthcare providers all over the world. Their employees need to be both fast and accurate in their work.

We will demonstrate the spreadsheet skills test that we administer in our hiring process.

4 - Spreadsheets Are Fine, But ...

Paul Rubin, Professor Emeritus, Michigan State University,

1312 Ramblewood Drive, East Lansing, MI, 48823,

United States of America, [email protected]

Spreadsheets play a significant role in operations research. They can be useful as data sources, for testing results, and for communicating those results to customers. Some analyses can be done entirely in spreadsheets, but often they are simply not the best tool for building and solving (and in some cases explaining) models. We will discuss pros and cons of using spreadsheets as a modeling platform.

TA16

16- West 207- CC

Combinatorial Auctions

Cluster: Tutorials

Invited Session

Chair: Subramanian Raghavan, Professor, University of Maryland,

Robert H. Smith School of Business, 4365 Van Munching Hall, College

Park, MD, 20742, United States of America, [email protected]umd.edu

1 - Combinatorial Auctions

Subramanian Raghavan, Professor, University of Maryland, Robert

H. Smith School of Business, 4365 Van Munching Hall, College

Park, MD, 20742, United States of America, [email protected]

This tutorial provides an introduction to combinatorial auctions. Topics discussed will include bidding languages (how bidders can bid), solution methods (how do you decide who wins what), and pricing (how much do the winners pay). Some recent innovations in auction formats and pricing will be discussed including the

Day-Raghavan procedure (also referred to as a core-selecting auction) which has been used in several government auctions to date. No familiarity with the area is assumed.

TA17

17- West 208 B- CC

Joint Session: INFORM-ED/SPRIG: Spreadsheet

Modeling: Past and Future?

Sponsor: INFORM-ED & Spreadsheet Productivity Research Interest

Group

Sponsored Session

Chair: Tom Groleau, Chair, Social Science Division, Carthage College,

2001 Alford Park Drive, Kenosha, WI, 53140, United States of America, [email protected]

1 - Creating VisiCalc: The First Spreadsheet

Daniel Fylstra, President, Frontline Systems Inc., P.O. Box 4288,

Incline Village, NV, 89450, United States of America, [email protected]

TA18

18- West 208 A- CC

Optimization Software in the Wild: Experiences From

Code Trenches

Sponsor: Optimization/Computational Optimization and Software

Sponsored Session

Chair: Cem Vardar, Senior Optimization Scientist, Revionics Inc.,

8700 East Via de Ventura, Suite 280, Scottsdale, AZ, 85258,

United States of America, [email protected]

1 - Using Unit and Integration Tests for Developing High Quality

Optimization Software

Cem Vardar, Senior Optimization Scientist, Revionics Inc.,

8700 East Via de Ventura, Suite 280, Scottsdale, AZ, 85258,

United States of America, [email protected]

Automated unit and integration testing of software is a widely used practice in

Software Engineering for ensuring high quality in code development. In this talk we are going to show some examples of how automated unit and integration testing is being used at Revionics in development of highly utilized operations research software. We’ll show examples of automated tests from an inventory allocation problem and a demand forecast modeling algorithm used in a price optimization setting.

2 - Column Generation in Practice: Efficiency Gains through Use of

Concurrency and Lessons Learned

Bertan Altuntas, Operations Research Software Developer,

LeanLogistics Inc., 1351 S Waverly Rd, Holland, MI, 49423,

United States of America, [email protected]

Column generation is a well-known technique used to solve large scale linear programming problems. At LeanLogistics, we have implemented a column generation based vehicle routing problem solver within our On-Demand

Transportation Management System. In this talk, I would like to focus on some practical aspects of implementing a column generation based solver, how we utilized parallel processing to improve efficiency of the solver and most importantly the lessons learned along the way.

266

INFORMS Phoenix – 2012

3 - “Cloud Manufacturing”– Distributed Caching on the

Semiconductor Manufacturing Floor

Patrick N Bless, Dr., Intel Corp., 5000 W. Chandler Blvd. Chandler,

CH5-255, Phoenix, AZ, 85048, United States of America, [email protected]

Semiconductor manufacturing equipment relies on large amounts of complex input data including recipes, test programs, and lot information. This data needs to be routed in real-time to process equipment on the floor to avoid idle time/utilization loss. Instead of isolating the equipment from each other via a centralized database system the author has developed a distributed, peer-aware in-memory data store designed to optimize data delivery rates while providing virtually unlimited scalability.

4 - Grid Computing in the Software as a Service

(SaaS) Environment

Jeff Moore, Revionics Inc., 8700 East Via de Ventura, Suite 280,

Scottsdale, AZ, 85258, United States of America, [email protected], Craig Morris

Retail planning applications such as pricing, promotions, and markdown optimization require weekly, daily, and on-demand processing to leverage the most up-to-date data and to enable retailers to realize greatest benefits. SaaS Grid

Computing shares resources among many retailers and provides much more efficient use of hardware. We will discuss lessons learned in building, running and maintaining a grid environment for optimization applications in a federated, multi-customer SaaS environment.

TA20

4 - Networks, Diversity and Stability: Cardiac Surgery Team

Structure and Performance

Emmanouil Avgerinos, PhD Candidate, University College London,

Department of Management Science and, Innovation,

Gower Street, London, WC1E 6BT, United Kingdom, [email protected], Bilal Gokpinar

The relationship between team structure and performance constitutes a matter of debate in the literature. In this study, we examine the relationship between network structure, team diversity and stability and their impact on efficiency and effectiveness of cardiac surgery operations. Using a large scale operational data from more than six thousand cardiac surgeries between 2005 and 2012, we highlight the significance of team structure on productivity and performance.

5 - Forecasting and Dynamic Adjustments of Staffing Level in

Hospital Operating Rooms

Su Xie, Stanford Graduate School of Business, 655 Knight Way,

Stanford, CA, 94306, United States of America, [email protected], Polly He, Stefanos Zenios

The staffing level in hospital operating rooms(time block of operating rooms assigned to a specialty) is decided biannually. We propose a model that hospital has the opportunity to forecast and adjust the staffing level dynamically as information of demand is updated over time. The optimal policy is proved to be state-dependent. Using the data from a teaching hospital, we are able to show that demand forecasts evolves as MMFE and conduct numerical experiment to study the value of adjustments.

TA19

19- West 211 A- CC

Workforce Considerations in Health Care

Contributed Session

Chair: Su Xie, Stanford Graduate School of Business, 655 Knight Way,

Stanford, CA, 94306, United States of America, [email protected]

1 - Combining Staffing with Quick-Response Decisions to Improve

Hospital Performance

Jan Schoenfelder, Doctoral Candidate, Indiana University, 1309 E.

Tenth Street, Bloomington, IN, 47405, United States of America, [email protected], Daniel Wright, Edwin Coe,

Kurt Bretthauer

We present a model that combines initial nurse staffing level decisions with two classes of quick-response decisions: (1) adjustments to the number of nurses in each unit, and (2) transfers of patients between units. Both improve hospital performance by anticipating or responding to day-to-day fluctuations in patient demand.

2 - A Comprehensive Bayesian Framework for Patient Aligned Care

Team (PACT) Work Load Estimation

Sara Shirinkam, Wayne state university, Wayne State University,

Detroit, MI, United States of America, [email protected],

Kai Yang, Adel Alaeddini

Patient Aligned Care Team (PACT) consists of a group of healthcare professionals with different expertise such as physician and nurse practitioner to provide accessible, comprehensive, and patient-centered care. We develop a comprehensive framework based on discrete cluster weighted modeling and

Bayesian inference to provide accurate, reliable and real-time estimate of PACT workload estimation. Using a real dataset from a medical center we demonstrate the performance of the proposed framework.

3 - WholeSurgeon Performance: Multiattribute Performance

Appraisals to Support Personnel Decisions

Robert Dees, University of Texas at Austin/US Army/Mayo Clinic,

18905 Colonial Manor Lane, Pflugerville, TX, 78660, United States of America, [email protected], Donald Potter, John Osborn,

Stephanie Heller

WholeSurgeon is a multiattribute model developed at the Mayo Clinic to measure the performance of surgeons. With this model, we are better able to communicate vision, mentor in line with our values, measure the talent of individuals, and support personnel decisions. In addition, WholeSurgeon provides a sound response variable for predictive recruiting models. We discuss elicitation from multiple experts, implementation, relationship to psychometrics, predictive techniques, and initial results.

TA20

20- West 211 B- CC

Issues In Public Health

Contributed Session

Chair: Serena Faruque, PhD Candidate, Stanford University,

330 Serra Mall, Stanford, CA, 94305, United States of America, [email protected]

1 - Calibrating Epidemiological Parameters of Influenza Simulation

Models in Short Decision Cycles

Diana Prieto, Assistant Professor, Western Michigan University,

Industrial Engineering, 1903 W. Michigan Ave., Kalamazoo, MI,

49008-5336, United States of America, [email protected],

Tapas Das

Influenza outbreak simulations are not yet adapted for operational response in progressing pandemic outbreaks. One of the features in need for operational adaptation is the model calibration. Usually, simulation models require human interaction to adjust the internal model parameters until obtaining a desired value. In this research, we propose a viral count driven approach to automate the calibration process and eliminate the need of the human component.

2 - Arrival Processes Based on SIR Models for Epidemic Diseases

Baykal Hafizoglu, Arizona State University, 699 S Mill Avenue,

Tempe, AZ, 85281, United States of America, [email protected],

Maria Rieders, Esma Gel, Lerzan Ormeci

Susceptible-Infected-Recovered models are commonly used to analyze the spread of epidemic, pandemic and infectious diseases. In this paper, we shed some light on the arrival process of patients at a hospital due to an epidemic. We modify a

Markovian SIR model by assuming that infected patients seek help at a hospital at an exponential rate. We present a recursive formulation for the calculation of the

Laplace transforms for the interarrival times for the sequence of patients seeking help.

3 - Evaluation of the Functional Result and Quality of Life after

Orthopedic Surgery

Chi-Chang Chang, Assistant Professor, Chung-Shan Medical

University Hospital, 110, Sec. 1, Chien-Kuo N. Rd., Taichung,

Taiwan-ROC, [email protected], Chen-Wen Chang

This research was to investigate the extent of psychological symptoms that individuals experience following orthopedic trauma and whether associated with quality of life. We adopted the SCL-90R scale and SF-36 to conducted Regression

Analyses and Structural Equation Modeling to predict the quality of life among study patients. The results showed psychological symptoms ultimately affect the quality of life.

267

TA21

4 - Optimization of Interventions for Reduction of Maternal

Mortality and Morbidity

Serena Faruque, PhD Candidate, Stanford University,

330 Serra Mall, Stanford, CA, 94305, United States of America, [email protected], Jeremy Goldhaber-Fiebert, Aparna Hegde

Approximately 350,000 women die yearly from complications of pregnancy or childbirth. WHO estimates that more than half of those deaths are preventable.

Providing preventative information - through people, or technology such as mobile phones - reduces maternal mortality and morbidity. We model how information improves patient adherence and health outcomes. Given a budget, mathematical optimization yields a package of technological and human intervention to maximize patient and societal outcomes.

5 - Modeling Hepatitis C Medications and Treatment Strategies

Marshall Kuypers, Sandia National Labs, 1515 Eubank SE,

Albuquerque, NM, United States of America, [email protected], Thomas Moore

A large cohort of patients infected with the Hepatitis C virus is progressing through the Veterans Hospital. New medications and treatment strategies have recently become available but there are many questions remaining regarding the optimal treatment. We present a system dynamics model that analyzes multiple facets of Hepatitis C disease treatment.

TA21

21- West 212 A- CC

CVaR and Chance Constrained Optimization in

Networks

Sponsor: Optimization/Networks

Sponsored Session

Chair: Baski Balasundaram, Assistant Professor, Oklahoma State

University, Industrial Engineering and Management, 322 Engineering

North, Stillwater, OK, 74074, United States of America, [email protected]

1 - Detecting Robust Cliques in Graphs with Uncertain Edge

Failures

Oleksandra Yezerska, PhD Candidate, Texas A&M University,

3131 TAMU, College Station, TX, 77843, United States of America, [email protected], Sergiy Butenko, Vladimir Boginski

We present an exact algorithm for detecting robust cliques in graphs subject to uncertain edge failures. As clique can be alternatively defined based on its various properties, the edge failure may yield different types of losses. We will discuss two loss functions (the sum and the max of vertex degree violations) and show that utilizing each of them ensures with certain probability the solution (after edge failures) will be s-defective clique or s-plex, respectively.

2 - Detecting r-robust 2-clubs in Graphs under Probabilistic

Edge Failures

Esmaeel Moradi, Oklahoma state University, 322 Engineering

North, Industrial Engineering & Management, Stillwater, OK,

74075, United States of America, [email protected],

Foad Mahdavi Pajouh, Baski Balasundaram

A subgraph of diameter at most k is called a k-club, a relaxation of cliques for k larger than 1. Extending this definition by requiring at least r distinct paths of length at most k, between all pairs of nodes results in the r-robust k-club model.

This talk will present preliminary results on a conditional value-at-risk (CVaR) based model to detect large r-robust 2-clubs in random graphs with probabilistic edge failures.

3 - On Chance-constrained Minimum Spanning K-core Problem under Probabilistic Edge Failure

Juan Ma, Research Assistant, Oklahoma State University,

92S University Place, Apt 6, Stillwater, OK, 74075, United States of

America, [email protected], Baski Balasundaram

Spanning k-cores can be used to design survivable networks that preserve low diameter conditions upon vertex deletion. This talk will focus on a chanceconstrained model of the minimum spanning k-core problem under probabilistic edge failures. We explored a branch-and-cut decomposition framework recently introduced in the literature, and this talk discusses problem specific enhancements via cutting-planes, scenario reduction approaches, and present preliminary numerical results.

INFORMS Phoenix – 2012

TA22

22- West 212 B- CC

40 Years of Forrest and Tomlin

Sponsor: Computing Society

Sponsored Session

Chair: John Tomlin, FICO, 181 Metro Plaza, San Jose, CA,

United States of America, [email protected]

1 - 40 Years of Forrest and Tomlin

John Forrest, Faster Coin, 19 Wix’s Lane, London, SW4 0AL,

United Kingdom, [email protected], John Tomlin

We review the the impact of the updating of triangular factors, and in particular our F-T method, on the speed of the simplex method, both at the time of publication 40 years ago, and in the years since on many mathematical programming (optimization) systems.

2 - A Review of Sparsity vs Stability in LU Updates

Michael Saunders, Stanford University, 475 Via Ortega,

Stanford, CA, United States of America, [email protected]

The Forrest-Tomlin update has stood the test of time within many generations of commercial mathematical programming systems. Its ease of implementation leads to high efficiency and evidently acceptable reliability. We review its relation to

Reid’s version of the Bartels-Golub update as implemented in LA05, LA15, and

LUSOL. In particular, we examine the extent to which FT implementations must

“live dangerously” in order to achieve the desired efficiency.

3 - Modern Implementation of the F-T method

Richard Laundy, FICO Xpress Development, United Kingdom, [email protected]

Although the original F-T method was designed to work efficiently with limited memory, out-of-core versions of the simplex method, it has proved suitable for modern in-core implementations. We outline some of these implementation details.

TA23

23- West 212 C- CC

Innovation in Analytics Award:

Semi-finalist Presentations I

Sponsor: Analytics

Sponsored Session

Chair: Michael Gorman, University of Dayton, Dayton, OH,

United States of America, [email protected]

1 - Robust Routing for Battery Electric Vehicles

Daniel Reich, Operations Research Analyst, Ford Motor Company,

2101 Village Road, MD 2122, Dearborn, MI, 48124, United States of America, [email protected], Mark Jennings, Mary Smith,

Michele Plattenberger, Dimitris Bertsimas, Thomas Magnanti,

Perry MacNeille, Ryan McGee, Ciro Soto, Joe Rork, Bill Frykman,

Susan Curry, Erica Klampfl, Matthew Fontana, Oleg Gusikhin

Battery electric vehicles (BEV) present many new challenges. Among them are limited driving range, long charging time, and highly variable energy consumption patterns. New vehicle electronics features are needed to help drivers make sound decisions: ones that avoid discharged battery situations and extend vehicle range. We are developing an innovative solution that combines traffic simulation, propulsion modeling, and optimization to provide BEV drivers with the best routes for their trips, based on time, travel preferences and energy consumption. By doing so, we aim to improve public confidence in vehicle electrification and achieve the social objectives obtainable through this emerging technology

2 - BracketOdds: Examining the NCAA Men’s Basketball

Tournament with Advanced Analytics

Sheldon Jacobson, Professor, University of Illinois, 201 N.

Goodwin Avenue (MC258), Urbana, IL, 61801, United States of

America, [email protected], Douglas King, Adrian J. Lee,

Alexander Nikolaev

The NCAA Men’s Basketball Championship draws considerable popular attention every March, often termed “March Madness”. Before this tournament begins, sports fans and neophytes attempt to predict a “perfect bracket” that correctly forecasts the winners of all tournament games, and submit their predictions to competitions (e.g., office pools, Internet-wide competitions). This research uses advanced analytics to assess the probability that a combination of seeds will advance in the tournament based on historical tournament results. Beginning in

2011, this model was adapted into a web-based tool, allowing users to assess prospective brackets online and leading to significant popular and media attention.

268

3 - Machine Learning for Power Grid Reliability: Predicting

Manhole Events in New York

Cynthia Rudin, Assistant Professor, Massachusetts Institute of

Technology, 77 Massachusetts Avenue, Cambridge, MA, United

States of America, [email protected], Steve Ierome, Delfina Isaac,

Rebecca Passonneau, Axinia Radeva

I will present innovations in proactive power grid maintenance stemming from a collaboration between Columbia University and Con Edison, NYC’s power utility company. Specifically, I will describe the Manhole Events Project, where the goal is to predict electrical failures, including manhole fires and explosions in New

York City. I will discuss the data mining process by which we transformed extremely raw historical data into a ranking model that predicts manhole vulnerability. Our ranked lists are currently assisting with the prioritization of future inspections and repairs in Manhattan, Brooklyn, and the Bronx.

TA24

24- West 213 A- CC

INFORMS Phoenix – 2012

Disease Monitoring and Optimal Screening and

Treatment Policies

Sponsor: Health Applications Society

Sponsored Session

Chair: Alireza Sabouri, University of British Columbia, Vancouver, BC,

Canada, [email protected]

1 - Optimal Hypertension Management for Risk-averse Patients with Scarce Resources

Gregg Schell, University of Michigan, 1205 Beal Ave., Ann Arbor,

MI, 48109, United States of America, [email protected],

Mariel Lavieri, Jeremy Sussman, Rodney Hayward

We develop a Markov decision process formulation to determine optimal hypertension treatment regimens for a risk-averse patient with scarce resources who seeks to minimize the expected number of coronary heart disease events over the planning horizon. Initial results from the model illustrate the diminishing returns of increased planning horizon resources. We also investigate the effect of risk-aversion, initial pretreatment blood pressure, and initial coronary heart disease risk.

2 - Evaluation of Breast Cancer Mammography Screening Policies

Considering Adherence Behavior

Mahboubeh Madadi, University of Arkansas,

4207 Bell Engineering Center 1, Fayetteville, AR, 72701,

United States of America, [email protected], Shengfan Zhang

The efficacy of mammography screening guidelines is highly associated with women’s compliance with these recommendations. However, none of the existing policies take women’s adherence behavior into consideration. Instead, perfect adherence is often assumed. In this study, a partially observable Markov model is proposed to evaluate and compare various screening policies, while incorporating variation in women’s adherence behavior.

3 - Personalized and Adaptive Diabetes Management: A Robust

Optimization Approach

Allison O’Hair, Massachusetts Institute of Technology, 77

Massachusetts Avenue, Bldg. E40-149, Cambridge, MA, 02139,

United States of America, [email protected], Dimitris Bertsimas

We propose an optimization approach to manage the blood glucose levels of type

II diabetes patients by optimizing food and exercise choices. We use robust integer optimization in a dynamic and adaptive way to determine food preferences and mixed-integer optimization to select an optimal meal and exercise plan that minimizes individual blood glucose levels. We have implemented an online software using the proposed approach and report evidence of its strength.

4 - Optimal Screening Strategies of Patients on the Kidney

Transplant Waiting List

Alireza Sabouri, University of British Columbia, Vancouver, BC,

Canada, [email protected], Steven Shechter, Tim Huh

The health condition of patients on the kidney transplant waiting list deteriorates while they are waiting for an organ arrival and hence they may no longer be suitable for transplant. Therefore, transplant centers screen waiting patients at various intervals to identify ineligible patients. We propose a model for finding screening strategies that minimizes the probability of offering a transplant to ineligible patients.

TA25

TA25

25- West 213 B- CC

Organ Transplantation Modeling

Sponsor: Health Applications Society

Sponsored Session

Chair: Ashley Davis, PhD Candidate, Northwestern University, 2145

Sheridan Road, Room C210, Evanston, IL, 60208, United States of

America, [email protected]

1 - An Empirical Analysis of the Effects of Kidney Allocation

Policies on Patient Behavior

William Williford, Graduate Student, Kellogg School of

Management, 2169 Campus Drive, Evanston, IL, 60208,

United States of America, [email protected],

Baris Ata, Che-Lin Su, Guenter Hitsch

We develop an optimal stopping model for patients on the deceased donor kidney wait list presented with kidney offers. Data provided by the United Network for

Organ Sharing (UNOS) is then used to estimate the model’s parameters via maximum likelihood estimation. The model is subsequently applied to several alternative kidney allocation policies in order to analyze their effects on the distribution of kidneys.

2 - Challenges in Training Cardio Thoracic Transplant Surgeons

Amy Cohn, University of Michigan, Ann Arbor, MI, United States of America, [email protected], Rishindra Reddy, Mark Daskin

Surgical residents training to become heart-lung transplant surgeons must perform a sufficient number of transplants during their residency/fellowship to become certified. There is an inherent conflict between the time required to do so

(along with the underlying stochasticity of transplant arrivals and durations)versus the regulations that limit trainees’ work hours. We quantify this conflict and propose alternative approaches.

3 - Random Effects ANOVA for Testing Heterogeneity of Brownian

Motion Cancer Growth

Gordon Hazen, Professor, Northwestern University, IEMS

Department, McCormick School of Enginering, Evanston, IL,

60208, United States of America, [email protected],

Neehar Parikh, Talia Baker, Daniel Apley

There is no consensus on the optimal treatment of liver cancer when transplant is an option. Deterministic growth models based on ODEs are unsuitable for decision analytic modeling where future growth is uncertain and growth is heterogeneous.

Instead, we fit geometric Brownian motion to natural history data for liver cancer, and adapt variance component models to test heterogeneity across data sources and patients. The fitted model allows for adaptive choice of treatment based on growth history.

4 - Improving Utilization and Outcomes of High Kidney Donor

Profile Index Kidney Transplants

Ashley Davis, PhD Candidate, Northwestern University, 2145

Sheridan Road, Room C210, Evanston, IL, 60208, United States of

America, [email protected], Sanjay Mehrotra,

Daniela Ladner, John Friedewald, Anton Skaro,

Jane Holl, Michael Abecassis

High Kidney Donor Profile Index kidneys (hKDPI) are most frequently discarded because of allocation system inefficiencies. We analyze current hKDPI kidney allocation and model three alternative allocation systems to compare each system’s expected allocation efficiency by mean kidney travel distance. We find that optimized directed Donor Service Area allocation reduces kidney travel distance by 42%. This reduction may potentially reduce hKDPI kidney wastage in a time of great organ shortage.

269

TA26

INFORMS Phoenix – 2012

TA26

26- North 221 A- CC

Healthcare Operations

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Rodney Parker, Associate Professor, University of Chicago, Booth

School of Business, 5807 South Woodlawn Ave, Chicago, IL, 60637,

United States of America, [email protected]

1 - Geographical Factors Deriving Deceased Donor Kidney

Procurement Rates

Mazhar Arikan, University of Kansas, School of Business,

Lawrence, KS, United States of America, [email protected],

Rodney Parker, Baris Ata

The deceased donor allocation system exists because of the severe shortage of available organs. Contrary to much of the current literature, we investigate the supply side. By using the donor registry data we show that the organ procurement rates vary significantly across different geographic areas and that the procurement rates are significantly correlated with the wait times. We build a structural model and conduct counterfactual analyses to understand the ways of increasing supply of organs.

2 - Physician Behavior and Patients’ Access to Healthcare with Payer Mix

Tingting Jiang, Student, Northwestern University, 2145 Sheridan

Road Room C210, Evanston, IL, 60208, United States of America, [email protected]

Across the U.S., Medicaid patients’ access to healthcare is diminishing. Low reimbursement rate compared with private health plan is usually blamed for physicians’ reluctance to accept Medicaid patients. We use a game theoretical model to study the physician’s capacity allocation and investment decisions as the competition for healthcare service intensifies. We find that physicians may prefer to provide service to patients with low and high reimbursement rates together due to scaled economy.

3 - On Hospice Operations under Medicare

Reimbursement Policies

Rodney Parker, Associate Professor, University of Chicago,

Booth School of Business, 5807 South Woodlawn Ave.,

Chicago, IL, 60637, United States of America, [email protected], Bradley Killaly, Tava Olsen,

Baris Ata

The Medicare’s hospice reimbursement policy consists of a daily payment for each patient under care with a global cap of revenues during the Medicare year, which increases for each newly admitted patient. We demonstrate several unintended consequences of the current reimbursement structure analytically. We also examine if the current policy is responsible for a recent spate of hospice bankruptcies. We propose an alternative policy, which overcomes the existing policy’s shortcomings

TA27

27- North 221 B- CC

New Business Models

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Karan Girotra, INSEAD, Boulevard de Constance, Fontainebleau,

France, [email protected]

1 - Resources for Results: Reducing the Funding Gap for Social

Development Projects

Milind Sohoni, Indian School of Business, Gachibowli, Hyderabad,

500032, India, [email protected], Sripad Devalkar

We study a market-based funding approach for social ventures wherein a not-forprofit intermediary solicits resources from donors for perfectly verifiable results post implementation. We analyze if such an approach reduces the gap between supply and demand of funds for social ventures. We answer the following questions: (i) Does an intermediary enhance welfare?, (ii) Is such a strategy sustainable and scalable?, and (iii) What is the impact of the intermediary’s operational costs?

2 - The Use (and Abuse) of Customer Voting Systems

Simone Marinesi, INSEAD, Boulevard de Constance,

Fontainebleau, 77305, France, [email protected],

Karan Girotra

This study examines a novel business process where firms conduct online polls asking customers to vote in favor of new product designs, and then uses the voting outcome to takes development and/or pricing decisions. Typically, customers are incentivized to vote by offering them purchase discount for the products they vote for. Our analysis highlights several pitfalls in the design of

Customer Voting Systems, and we provide prescription on how such systems should be employed.

3 - The Benefits of Decentralized Decision Making in

Supply Chains

Elena Belavina, Assistant Professor, Chicago Booth School of

Business, 5807 South Woodlawn Avenue, Chicago, IL, 60637,

United States of America, [email protected],

Karan Girotra

The inefficiency of decentralized decision making is one of the most influential findings of the supply chain coordination literature. We show that with the possibility of repeated trade, decentralization can be beneficial in improving supply chain performance. With decentralized decision making, it is easier to incentivize players to coordinate on efficient actions. It is manifested both in higher supply chain profits and more flexibility in contracting terms that are acceptable to all players.

4 - The Impact of Workforce Management Software Introduction upon Performance

Serguei Netessine, Professor, INSEAD, Boulevard de Constance,

Fountainbleau, 77305, France, [email protected],

Tom Tan

We analyze the impact of introducing workforce management software which induces competition among restaurant servers. We show that the tournament can be an effective tool in incentivizing workers.

TA28

28- North 221 C- CC

Empirical Healthcare Operations Management

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Diwas KC, Emory University, 1300 Clifton Road, Atlanta, GA,

30322, United States of America, [email protected]

1 - Revisiting the NH Needle Exchange Program: A Sensitivity Test for Seviations from Uniform Sampling

Donald Lee, Yale University, 135 Prospect St., New Haven, CT, [email protected]

Program evaluation relies critically on the quality of and methodology for outcome measurement. However, it is often difficult to obtain information on outcomes for the entire population, or even a representative sample. We introduce a sensitivity test for deviations from equal probability sampling of units, and use it to test the robustness of the reported outcomes in the New Haven needle exchange program. Joint work with Peter Aronow (Yale Political Science).

2 - Patient Revisit Interval in Primary Care: Impact on Patient

Health and Panel Size

Hessam Bavafa, Doctoral Candidate, The Wharton School, UPenn,

3730 Walnut St., Hunstman Hall, Suite 500, Philadelphia, PA,

19104, United States of America, [email protected],

Christian Terwiesch, Sergei Savin

We study the decision made by primary care physicians regarding patient revisit interval (RVI). We provide insights on the impact of RVI on panel size, emergency visits and patient health. We also evaluate the value of remote patient monitoring with which the physician receives informative signals from the patient such as blood pressure, weight, etc. between patient visits. We show that utilizing these signals can help the physician increase his panel size and quality of care.

3 - Efficiency Analysis of Hospitals

Sriram Venkatraman, Doctoral Student, Clemson University, 100

Sirrine Hall, Clemson, SC, 29634-130, United States of America, [email protected], James Sun, Maria Ibanez, Aleda Roth,

Anita Tucker

This paper estimates efficiencies of hospitals in the U.S., using input and output measures from publically available data sources. We introduce a consistent bootstrap technique from data envelopment analysis literature to compare hospitals on various bases and draw meaningful statistical inference. We also test the assumption of returns to scale.

4 - Does the Individual Mandate Free You To Choose

Diwas KC, Emory University, 1300 Clifton Road, Atlanta, GA,

30322, United States of America, [email protected],

Sriram Venkataraman

We study a natural policy experiment - the Massachusetts healthcare reform law to examine the impact of universal healthcare on patient volume and the allocation of volume of previously uninsured patients across different emergency departments. Our model of patient choice finds that post-policy, utility maximizing patients choose to avoid safety-net hospitals.

270

INFORMS Phoenix – 2012

TA29

29- North 222 A- CC

Scheduling and Capacity Management in Hospitals

Sponsor: Manufacturing & Service Oper Mgmt/

Healthcare Operations/SIG

Sponsored Session

Chair: Nan Liu, Columbia University, 600 W 168th St., Room 603,

New York, NY, 10032, United States of America, [email protected]

1 - Procedure Room Allocation Decisions with Demand Uncertainty

Bjorn Berg, George Mason University, 4400 University Drive,

MS 4A6, Fairfax, VA, 22030, United States of America, [email protected], Brian Denton

Urgent add-on procedures make planning and scheduling decisions difficult in outpatient procedure center settings. The decisions of dynamically allocating procedures to procedure rooms are modeled as a stochastic online bin-packing problem. A special case is formulated as multi-stage stochastic integer program.

Results for optimal capacity management decisions are presented as well as results for easy-to-implement heuristics.

2 - On Quantifying the Benefits of Coordinating OR

Capacity Management

Fei Li, University of Minnesota, 111 Church Street S. E.,

Minneapolis, MN, 55455, United States of America, [email protected], Diwakar Gupta

Patient Protection and Affordable Care Act is intended to align health care providers’ and suppliers’ incentives to provide patients the right care at the right time. Payment mechanisms will reward providers for better health outcomes, rather than for the volume of services provided. We develop models to quantify the benefits that may accrue from more efficient use of OR capacity when hospital and physician incentives are aligned. These models are tested on sample data from a community hospital.

3 - Multi-resource Allocation Scheduling of Elective and

Emergency Surgery in Dynamic Environments

Van-Anh Truong, Columbia University, [email protected],

Tim Huh, Nan Liu

We formulate the first model of allocation scheduling for emergency and elective surgeries in a dynamic environment, where distributional information about demand and resource availability is continually updated and scheduling decisions exploit new information as it arises. We allow multiple resources to be considered, and patients to renege.

4 - Capacity Planning for Cancer Prevention

Aaron Ratcliffe, [email protected],

Ann Marucheck, Wendell Gilland

We develop a queuing network to optimize providers’ capacity and pricing decisions for consumers seeking repeat service. Consumers’ repeat arrival behavior is a function of guideline compliance and the expected value, price, and wait-time for service. Utilizing public health data for colorectal cancer screening, we inform government planning and intervention.

TA30

30- North 222 B- CC

Risk Management in Agricultural Markets

Sponsor: Manufacturing & Service Oper Mgmt/iFORM

Sponsored Session

Chair: Onur Boyabatli, Singapore Management University,

50 Stamford Road, Singapore, Singapore, [email protected]

1 - Yield Management in Agricultural Markets

Kwan Eng Wee, Assistant Professor, Singapore Management

Univesity, 50 Stamford Road #04-01, Singapore, 178899,

Singapore, [email protected], Onur Boyabatli

This paper considers a processor who has to decide on the sourcing options for an agricultural input in the face of farm yield uncertainty of that input. The input can be sourced from farmers on a long-term contract basis in advance, or through an open market on the day. The processors can also use different pricing schemes on the output side to hedge against the yield uncertainty. Our work studies the impacts of these options on the processor’s operations and performance.

2 - Agricultural Planning of Annual Plants under Demand,

Maturation, Harvest, and Yield Risk

Baris Tan, Professor of Operations Management, Koc University,

Rumelifeneri Yolu, Sariyer, Istanbul, 34450, Turkey, [email protected]

We present a planning methodology for a firm whose objective is to match the random supply of annual premium fruits and vegetables from a number of

TA31

contracted farms and the random demand from the retailers during the planning period. The proposed method determines the farm areas and the seeding times in such a way that the expected total profit is maximized. The profit obtained by this method is substantially higher compared to the planning approaches where the uncertainties are not considered.

3 - Dynamic Risk Management of Commodity Operations

Sripad Devalkar, Indian School of Business, Gachibowli,

Hyderabad, 500032, India, [email protected],

Ravi Anupindi, Amitabh Sinha

We model the dynamic risk management problem for a commodity processor in a multi-period setting. We use a time-consistent risk measure based on the conditional value at risk (CVaR) and obtain the optimal operational and financial hedging policy in a partially complete market. We quantify the value of timeconsistent decision making in terms of mean-risk tradeoff. We also quantify the value of excess procurement and processing capacities as operational hedging levers.

4 - A Dynamic Mechanism for Achieving Sustainable Quality Supply

Fang Liu, Nanyang Technological University, 50 Nanyang Avenue,

Singapore, 639798, Singapore, [email protected],

Jing-Sheng Song, Tracy Lewis

Various companies have realized the importance of sustainable quality supply and initiated sustainability guidelines. However, are these guidelines sufficient to ensure sustainable quality supply? We design a dynamic mechanism that implements sustainable quality supply. Part of our mechanism suggests a two-part nonlinear tariff payment which is consistent with the existing industry guidelines; whereas the other part suggests new elements such as the probabilistic decision rights.

TA31

31- North 222 C- CC

Pricing and Services

Sponsor: Manufacturing & Service Oper Mgmt/Service Operations

Sponsored Session

Chair: Laurens Debo, University of Chicago, Chicago, IL,

United States of America, [email protected]

1 - Markdown Management: Pricing as a Signaling Device

Gad Allon, Northwestern University, 2001 Sheridan Rd, Evanston,

IL, United States of America, [email protected],

Achal Bassamboo, Ramandeep Randhawa

We study the classical revenue management scenario of a monopolist retailer selling to customers over a finite horizon. The retailer possesses information about the aggregate demand, whereas customers have private information on their valuations and are strategic. We study how the profit maximizing retailer strategically uses price as a signaling device.

2 - Pricing and Operational Performance in Discretionary Services

Raj Rajagopalan, University of Southern California, Marshall

School of Business, Los Angeles CA, United States of America

[email protected], Chunyang Tong

In discretionary services, the value or quality derived by a customer from a service depends upon the time the service provider (SP) devotes to the customer and the valuation differs across customers. We identify the optimal pricing scheme and explore the impact of two widely used pricing schemes, time-based billing and fixed fee, on many dimensions of service performance including SP’s profitability, demand, utilization, and congestion level.

3 - Selling to the Censored Newsvendor: Optimal Wholesale Price with Demand Learning

Yunru Han, Columbia University, 3022 Broadway, New York, NY,

10027, United States of America, [email protected],

Fangruo Chen

One classic result concludes that the censored newsvendor sets a higher inventory level for demand learning instead of optimizing one-period profit. Yet if the supplier can change the wholesale price, she may either exploit or collaborate on the retailer’s learning policy, depending on the contracting and information sharing process. We study the supplier’s optimal wholesale price in various situations, and compare the profitability and efficiency of active learning and myopic policies under it.

4 - Queue Joining Strategies when Expected Service Time and

Value are Unknown

Laurens Debo, University of Chicago, Chicago, IL, United States of

America, [email protected], Senthil Veeraraghavan

When the expected service time and service value are not perfectly known to a consumer base, but, positively correlated, both posterior value and waiting cost increase in the queue length. We show that this leads to non-monotone queue joining strategies.

271

TA32

INFORMS Phoenix – 2012

TA32

32- North 223- CC

Economics of Supply Chain Management

Sponsor: Manufacturing & Service Oper Mgmt/Supply Chain

Sponsored Session

Chair: Manu Goyal, David Eccles School of Business,

University of Utah, Salt Lake City, United States of America,

[email protected]

1 - Stage-Gate Contracts for New Product Development

Chunlin Wang, David Eccles School of Business, University of

Utah, Salt Lake City, [email protected],

Krishnan Anand, Glen Schmidt

We gain new insights into why a firm might use a stage-gate contract when outsourcing New Product Development. The contract helps address both adverseselection and moral hazard issues. We study the optimal number and placement of gates, and compare the performance of stage-gate contracts with traditional outsourcing contracts.

2 - Contracting for On-Time Delivery in the US Influenza Vaccine

Supply Chain

Soo-Haeng Cho, Carnegie Mellon University, 5000 Forbes Avenue,

Pittsburgh, PA, 15213, United States of America, [email protected], Tinglong Dai, Fuqiang Zhang

Motivated by the flu vaccine industry, we consider a supply chain that involves three sources of uncertainties: design, delivery, and demand. In such a supply chain it is critical for the firms to choose the right production quantity at the right time. It is shown that several well-known contracts fail to coordinate the supply chain. So we propose new coordinating contracts that are reported in practice but not studied in the extant literature.

3 - Pricing and Rationing When Selling to Snobbish Consumers

Kenan Arifoglu, Assistant Professor, University College London,

Gower Street, London, WC1E 6BT, United Kingdom, [email protected], Seyed Iravani, Sarang Deo

We study the pricing policy and capacity decision of a firm selling snob-appeal products such as high-fashion luxury goods to forward-looking, exclusivityseeking (snobbish) consumers. We use our model to explain the heterogeneity in pricing of snob-appeal products; some firms adopt uniform pricing while others mark down. We also find cases, where firms create scarcity and mark the price down simultaneously during the selling season.

4 - Honesty as an Optimal Policy under Bounded Rationality

He Chen, R.H. Smith School of Business, University of Maryland,

College Park, United States of America, [email protected]d.edu,

Manu Goyal, Krishnan Anand

Incomplete contracts are vulnerable to opportunism. In a multi-period relationship with incomplete contracts between a boundedly-rational manufacturer and his supplier, we prove that an ‘honest’ manufacturer can outperform a quintessential ‘opportunistic’ manufacturer, even though the opportunistic manufacturer can mimic the honest manufacturer. Thus, honesty emerges endogenously as an optimal policy.

TA33

33- North 224 A- CC

Inventory Review Policies

Contributed Session

Chair: Raik Özsen, University of Cologne, Albertus-Magnus-Platz,

Cologne, 50923, Germany, [email protected]

1 - A New Approach to Inventory Management

Rasoul Haji, Professor, Sharif University of Technology, Azadi Ave.,

Tehran, Iran, [email protected]

We introduce a new approach to inventory management called No Review

Inventory Management (NORIM) which is simple, easy to apply, and eliminates the suppliers’ demand uncertainty and safety stock. For a special case we obtain the optimal total cost rate and compare it with the (s, S) policy optimal solution.

2 - Generalized EOQ Model with Two Levels of Warehouse Storage

Maxim Bushuev, PhD Candidate, Kent State University,

540 S Water St., #311, Kent, OH, 44240, United States of America, [email protected]

A continuous review (Q, R) model with backorder is extended to a model with two levels of storage (owned and rented warehouses) and alternative cost dimensions such as ([$/unit/unit time], [$/unit time], and [$/unit]). Optimal conditions are derived and the convexity is discussed.

3 - Ranking the Quality of Lead Time Demand Models

Yasin Unlu, Research Scientist, LLamasoft, Inc., 3404 Columbus

Lane, Ann Arbor, MI, 48103, United States of America, [email protected], Manuel D. Rossetti

This paper discusses a number of different measures to rank the quality of given

LTD models for the classic continuous review (r, Q) inventory system. Two types of measures are used: 1) test measures; namely, chi-square, Kolmogorov-Smirnov and sum of squared error that are estimated based on utilizing the observed LTD values during simulation, 2) error measures that are estimated based on a number of inventory performance measures.

4 - A Generalized Basestock Inventory Policy

Abhilasha Aswal, International Institute of Information

Technology Bangalore, 26/C Electronics City, Bangalore, India, [email protected], G N Srinivasa Prasanna

We present a generalization of the basestock policies for multi-item inventory optimization. Co-related inventory constraints can form arbitrary polytopes and may have many faces. We can have a generalized basestock policy, with an outer re-order polytope and an inner trigger polytope. In contrast to a standard (s,S) policy, here the threshold and reorder point of each product keep changing, as a function of available inventory of the other products.

5 - An Optimal Expediting Policy for Periodic-review

Inventory Systems

Raik Özsen, University of Cologne, Albertus-Magnus-Platz,

Cologne, 50923, Germany, [email protected],

Ulrich Thonemann

Most inventory models assume that lead times are exogenously given. However, in many situations inventory managers have the option to expedite open orders.

We incorporate the expediting option into a periodic review inventory model and solve the model optimally. Unlike previous research, we evaluate the objective function exactly and show how the optimal solution can be computed efficiently.

We estimate the benefit of order expediting based on real data from a global equipment manufacturer.

TA34

34- North 224 B- CC

Experimental Research on Innovation/NPD

Cluster: New Product Development

Invited Session

Chair: Jeremy Hutchison-Krupat, Darden School of Business,

University of Virginia, 100 Darden Blvd, Charlottesville, VA, 22903,

United States of America, [email protected]

1 - Leaps in Innovation

Joel Wooten, University of Pennsylvania, The Wharton School,

Philadelphia, PA, United States of America, [email protected]

Experiments have the ability to test theory and validate hypotheses. Here, we look at algorithmic innovation tournaments from Kaggle.com to examine the effect of incremental progress in innovation. We use submission-level entries in data-driven tournaments to examine whether the innovation horizon inspires additional innovation.

2 - Ambiguous Problem Complexity, Group Synergy and

Performance: An Experiment

Elliot Bendoly, Associate Professor, Emory University,

1300 Clifton Road, Atlanta, GA, 30322, United States of America, [email protected], Stylianos Kavadias,

Svenja Sommer

Kavadias and Sommer (2009) describe a normative model predicting strong collaborative group problem solving in relatively simple settings, but weakness in increasingly complex ones. We revisit the issue through a computer-based experiment. The value of the normative model is strengthened through the consideration of member diversity, focus, control and accountability.

3 - Tolerance for Failure and Incentives for Collaborative Innovation

Jeremy Hutchison-Krupat, Darden School of Business, University of Virginia, 100 Darden Blvd, Charlottesville, VA, 22903, United

States of America, [email protected], Raul Chao

An organization’s ability to innovate can be enhanced by managing risk-taking behavior through monetary incentive schemes and through an organizational culture that tolerates failure. This paper reports the results of two controlled experiments aimed at understanding how tolerance for failure and incentives impact the decisions of individuals engaged in a collaborative innovation initiative.

272

TA35

35- North 225 A- CC

Behavioral Strategies

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: John Wilson, Ivey School of Business, Univ. of Western Ontario,

London, Canada, [email protected]

1 - New Results Concerning Probability Distributions with

Increasing Generalized Faiure Rates

Prakash Mirchandani, University of Pittsburgh, Katz Graduate

School of Business, Pittsburgh, PA, 15260, United States of

America, [email protected], Mihai Banciu

The generalized failure rate (GFR) of a random variable is an extension of its hazard rate. If the valuation distribution of a product has an (strictly) increasing

GFR, then the associated revenue function is unimodal, and its global maximum is (uniquely) specified. This talk proposes an alternate economic viewpoint that extends the commonly accepted definition of the GFR and presents some new properties related to the closure of IGFR distributions under different operations.

2 - Competing with Customer Returns Policies

Jing Chen, Associate Professor, Faculty of Business & Economics,

University of Winnipeg, 460 Portage Avenue, Winnipeg R3B 2E9

CANADA, Winnipeg, MB, R3B 2E9, Canada, [email protected], Bintong Chen

Customer returns policies vary in the retailing industry. We develop a model to illustrate how retailers with different customer returns policies compete for customers via pricing. We further extend the model to demonstrate how customer returns policy can be used as a competing strategy to acquire customers and its associated impact on pricing and market segmentation.

3 - Behavioral Bidding Strategies

Fredrik Odegaard, Assistant Professor, Richard Ivey School of

Business, 1151 Richmond St. N, London, On, N6G4X3, Canada, [email protected]

Auctions and bidding strategies have a rich and extensive research history. While theoretical results regarding optimal bidding strategies are well-established, there are also numerous empirical anomalies. In this paper we investigate certain simple bidding heuristics and compare those with observed online bidding behavior.

4 - A Behavioral Perspective of Strategic Interactions in Revenue

Management Markets

Srinivas Krishnamoorthy, University of Western Ontario, 1151

Richmond Street N, Room 2R14A, London, ON, N6A 3K7,

Canada, [email protected]

Aggregate pricing decisions of travel capacity providers often appear to be economically irrational. A typical example is a price war between competing airlines or the special deals offered by competing hotels. Standard game theoretic models fail to explain these decisions. We use behavioral game theory to obtain more realistic predictions of provider behavior in competitive revenue management markets.

5 - Block T Discounts on the Internet

John Wilson, Ivey School of Business, Univ. of Western Ontario,

London, Canada, [email protected], Jing Chen

Many of the new internet discount outlets have some form of a promotion involving T items offered at a discount price. Selling at a discount would cannibalise from the retailers current customers since they could buy at the cheaper price. However, it would take customers from the retailer’s competitors.

For a wide class of distributions, we derive necessary and sufficient conditions for a promotion to be optimal.

INFORMS Phoenix – 2012

TA37

TA36

36- North 225 B- CC

Topics in Pricing and Revenue Management

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: N. Bora Keskin, University of Chicago, Booth School of Business,

5807 South Woodlawn Avenue, Chicago, IL, 60637,

United States of America, [email protected]

1 - Pricing Managed Lanes

Caner Gocmen, Columbia Business School, 3022 Broadway,

Uris Hall, 4V, New York, NY, 10027, United States of America, [email protected], Robert Phillips

A managed lane is loosely defined as a separated highway lane operated with active flow management. The entity responsible from operating the lane manages the flow to the lane by levying tolls or setting regulations. Managed lanes with tolls are becoming an important way to finance new public works. We investigate the optimal operating policies for managed lanes with tolls by taking the viewpoint of the operating entity which may seek to maximize the revenue or the social surplus.

2 - Dynamic Matching with Heterogeneous Goods

Hua Zheng, PhD Candidate, Columbia University,

3022 Broadway, New York, NY, 10027, United States of America, [email protected], Ciamac Moallemi, Costis Maglaras

We study a dynamic market for heterogenous goods: buyers and sellers arrive dynamically over time; sellers supply heterogenous units to the market; buyers are endowed with a choice model to select among heterogeneous alternatives; participants have idiosyncratic valuations and incur delay costs while waiting to trade. We study the equilibrium of such a market, and characterize its “depth” of active sellers and buyers as well as price dispersion. A motivating application is residential real-estate.

3 - Bounded Rationality in Strategic Interactions in Supply Chain

Georgia Perakis, Massachusetts Institute of Technology, Cambridge,

MA, United States of America, [email protected], Basak Kalkanci

Despite the standard assumption of perfect rationality in theoretical models, decision makers are often boundedly rational and are subject to errors and biases, particularly in interactions with other decision makers. We examine how bounded rationality drives the interaction of multiple players in supply chain to find out (1) how much is lost by assuming perfect rationality and (2) whether bounded rationality can work as an advantage for a decision maker.

4 - Dynamic Pricing with Demand Learning in a

Changing Environment

N. Bora Keskin, University of Chicago Booth School of Business,

5807 South Woodlawn Avenue, Chicago, IL, 60637, United States of America, [email protected], J. Michael Harrison,

Assaf Zeevi

We consider a dynamic pricing problem in which a seller faces an unknown demand model that can change over time. We evaluate several pricing policies via simulation, and provide simple guidelines to implement price experimentation while keeping track of changing market conditions.

TA37

37- North 226 A- CC

Network Interdiction and Defense Modeling

Sponsor: Location Analysis

Sponsored Session

Chair: Mark Daskin, Professor, University of Michigan,

Department of IOE, 1205 Beal Avenue, Ann Arbor, MI, 48109,

United States of America, [email protected]

1 - An “All Hazards” View of Supply Chain Resilience

David Alderson, Operations Research Department, Naval

Postgraduate School, Monterey, CA, United States of America, [email protected], Marco Laumanns

We model tactical planning in a supply chain that must mitigate and balance protection against both “worst-case” disruptions, as might be caused by the deliberate actions of an intelligent adversary, and “most likely” disruptions as caused by non-deliberate hazards, such as natural disasters or technological failures.

273

TA38

INFORMS Phoenix – 2012

2 - A Heuristic Model for Highway Interdiction

Tolga Seyhan, Operations Research Scientist, Amazon.com,

345 Boren Ave. N, Seattle, WA, 98109, United States of America, [email protected], Lawrence V. Snyder

We consider a network interdiction model in which a smuggler uses a network of roads and an interdictor deploys patrol units to increase the smuggler’s cost of using particular roads. We heuristically reformulate the smuggler’s problem, which is a prize-collecting Steiner tree problem, and solve the interdictor’s problem as a single-level mixed-integer program.

3 - Defender-interdictor Location Models with Uncertain Futures

Kaiyue Zheng, IOE Department, University of Michigan,

1205 Beal Avenue, Ann Arbor, MI, 48109,

United States of America, [email protected], Mark Daskin

We model scenario-based two-stage p-robust defender-interdictor problems. The defender allocates resources to mitigate the interdictor’s actions, while the interdictor wants to maximize damage while inflicting at least a minimal level of damage in each scenario. We present two models with different interdictor options. Algorithms and computational results are given.

TA38

38- North 226 B- CC

Supply Chain Services: Models and Applications – I

Sponsor: Service Science

Sponsored Session

Chair: M. Ali Ulku, Assistant Professor, School of Management and

Leadership, Capital University, 1 College and Main, Columbus, OH,

43209, United States of America, [email protected]

1 - Impact of Inventory Control Policies on On-time Delivery in a

Divergent Supply Chain

Joon Young Lee, Graduate Student, ASU ME, 850 S. McAllister

Ave. ISTB2 275, Tempe, AZ, 85287, United States of America, [email protected], Seung Hwan Kim

Assignment and placement policies (rationing and soft-pegging) are considered when a customer order is materialized at a diversification point in a divergent supply chain in order to maximize customer service level. A semiconductor wafer fab is modeled. We experiment with various input scenarios to see the effects of our policies. We show that on-time delivery is maximized in equilibrium state at lot bank by proper assignment policies, and lot bank is selected as a good location.

2 - Integrated vs. Dedicated: Delivery System Choice for a Vertically

Differentiated Product Line

Aditya Jain, Assistant Professor, Indian School of Business,

Gachibowli, Hyderabad, India, [email protected], Ram Bala

We analyze service systems choice – integrated or dedicated, for supporting services, which are provided along with vertically differentiated products. We characterize the preferred choice as a function of performance deterioration that may result from mix variability. We show the impact of market cannibalization on strategic operations decision.Further, this impact of cannibalization varies depending on the market scale as well as market mix.

3 - Simulation Study of a Three-level CPFR System

James Bookbinder, Professor, University of Waterloo, 200

University Ave. West, CPH 4323, Waterloo, ON, N2L 3G1, Canada, [email protected], Amanda Cha

Sales data on several product groups of a well known manufacturer are employed.

That firm participates in “Collaborative Planning, Forecasting and Replenishment” with a major retail chain. Through a stochastic simulation model, we contrast

CPFR to Vendor Managed Inventory, and to Independent Sourcing. Comparisons are made on total system cost, aggregate inventory levels, and fill rates.

4 - Fair Supply Chains: Implications and Complications of Fair

Trade in Supply Chain Modeling

M. Ali Ulku, Assistant Professor, School of Management and

Leadership, Capital University, 1 College and Main, Columbus, OH,

43209, United States of America, [email protected]

Fair Supply Chains (FSCs) aim to achieve a greater equity in international trading, and thus, sustainable development. As consumers become more socially responsible, the demand and therefore the shelves for certified fair trade goods in big retailers have been expanding rapidly. In many aspects, FSCs greatly differ from Commercial Supply Chains (CSCs). This research introduces modeling challenges in FSCs and proposes a fairness-coordination mechanism that can be integrated to a global CSC.

TA39

39- North 226 C- CC

Supply Chain in Retail Management

Contributed Session

Chair: Arnd Huchzermeier, Professor, WHU - Otto Beisheim School of

Management, Burgplatz 2, Vallendar, 56179, Germany, [email protected]

1 - Impacts of Retailer’s In-Store Marketing Effort in Private Label and National Brand Competition

Tieming Liu, Associate Professor, Oklahoma State University, 322

Engineering North, Industrial Eng and Management, Stillwater,

OK, 74078, United States of America, [email protected],

Yingjue Zhou, Gangshu Cai

This paper examines a supply chain with a retailer selling a private label product

(PL) and a national brand product (NB). We assume the retailer pays in-store marketing effort to promote sales. We study the retailer and the national brand manufacturer’s pricing and marketing effort decisions. We show that under certain conditions, the retailer may assign the in-store marketing effort to NB, and the NB manufacturer may benefit from the introduction of PL.

2 - Selecting Store Delivery Patterns in Grocery Retailing

Michael Sternbeck, Catholic University Eichstaett Ingolstadt,

Auf der Schanz 49, Ingolstadt, 85049, Germany, [email protected], Heinrich Kuhn

We consider a grocery retail chain consisting of the three subsystems warehousing, transportation and instore-logistics. In this environment, determining store-specific delivery patterns is considered as an important interdependent problem in internal supply chain planning on a tactical level. We use an optimization approach to select store delivery patterns. The results show that integrating instore logistics is of major importance.

3 - Optimal Policies for Recovering the Value of Consumer Returns

Paolo Letizia, Assistant Professor, Erasmus University, Rotterdam

School of Management, Rotterdam, Netherlands, [email protected]

This paper characterizes a Pareto optimal returns policy between a manufacturer and a retailer who receives consumer returns. The manufacturer may take a costly hidden action that reduces the expected number of products returned by consumers, which when realized is hidden information known only to the retailer. When faced with consumer returns, the retailer must decide whether to send the product back to the manufacturer, who harvests a low salvage value, or to engage in costly refurbishment that permits the returned product to be resold to consumers. We find that the optimal returns policy may be implemented through the payment by the manufacturer of a full refund to the retailer of the wholesale price for any returns, as well as a bonus paid to the retailer that is decreasing in the number of returns to the manufacturer.

4 - Returns Management and Secondary Markets

Arnd Huchzermeier, Professor, WHU-Otto Beisheim School of

Management, Burgplatz 2, Vallendar, 56179, Germany, [email protected], David Schroeder

This presentation analyzes the optimal returns management policy of online retailers under competition and in the presence of a secondary market. A big data set of a pan European online retailer of fashion products — offering a free returns policy of up to 100 days after a customer’s purchase — is utilized to provide new insights on actual customer behaviour and effective returns policies.

TA40

40- North 227 A- CC

Sustainable Practices and Innovations

Sponsor: Energy, Natural Res & the Envi/ Environment and

Sustainability

Sponsored Session

Chair: Xu (Cissy) Yang, Postdoctoral Associate, Massachusetts Institute of Technology, 1 Amherst Street, E40-211, Cambridge, MA, 02142,

United States of America, [email protected]

1 - Root Causes of Product Expiration in the Consumer Packaged

Good Industry

Arzum Emel Akkas, Massachusetts Institute of Technology,

77 Massachusetts Avenue, Cambridge, MA, 02139,

United States of America, [email protected], David Simchi-Levi

Our work investigates the root causes of product expiration at retail stores using archival operations data from a large beverage company. Using binomial regression, we quantify the impact of product configuration, warehouse aging of products, warehouse congestion, inventory policy, and suboptimal shelf arrangements on product expiration observed at retail stores.

274

INFORMS Phoenix – 2012

TA42

2 - A Multiple-player-game Approach to Agricultural Water Use in

Seasonal Drought Area: A Case in Brazil

Zhouyang Lu, Duke University, 121 Hudson Hall CEE, Box 90287,

Durham, NC, 27708-0287, United States of America, [email protected]

Seasonally dry areas, widely distributed in the world, are facing an intensive disparity between the lack of water and the great demand of social development.

This work discusses the guarantee of micro-irrigation running in dry season in

Fazenda Tamandua. The effects of the landlord’s supervision on the microirrigation are analyzed with a multiple-player game model. The outputs from the model and data show the influence of landlord’s supervision on the moral hazard between the stakeholders.

3 - Modeling Multiple Objectives in Green Building Decisionmaking

John Dickson, University of Texas at Arlington, Arlington, TX,

United States of America, [email protected], Victoria Chen

Green building seeks to improve building sustainability. A decision-making framework is presented that utilizes design and analysis of computer experiments to study the impact of building options on sustainability objectives. The QUick

Energy Simulation Tool (eQUEST) is employed to evaluate building performance.

Modeling of multiple objectives is achieved using treed regression and treed multivariate adaptive regression splines with the seemingly unrelated regressions method.

4 - Collaborative Planning in Closed-loop Supply Chains

Aman Gupta, Assistant Professor, Embry-Riddle Aeronautical

University-WorldWide, 10733 Copper Ridge Dr, Louisville, KY,

40241, United States of America, [email protected]

Collaborative planning involves information sharing and coordination of the planning tasks of independent supply chain members. Collaboration or information sharing can be at different levels including, minimal, deep involvement, and somewhere in between. The research presents an overview of the literature published on impact of collaborative planning in managing closedloop supply chains and potential future research in this area.

TA41

41- North 227 B- CC

Forestry Session II

Sponsor: Energy, Natural Res & the Environment/Forestry

Sponsored Session

Chair: Marc McDill, Associate Professor, Pennsylvania State School of

Forest Resources, 310 Forest Resources Building, University Park, PA,

16802, United States of America, [email protected]

1 - Fire, Water and Owl: Multi-objective Forest Planning in the

Deschutes National Forest, USA

Sandor Toth, Assistant Professor, University of Washington, School of Environmental and Forest Scien, Seattle, WA, 98195,

United States of America, [email protected], Svetlana Kushch,

Thomas Mafera, Leo Yanez, Gregory Ettl, Robert Deal

We present a multi-objective optimization application in a municipal watershed in the Deschutes National Forest, where water quality, spotted owl habitat, and fire risk mitigation are all critical but competing services. For example, fuel treatments increase sediment in the short run, but in the long run the risk of wildfires will be lower and sedimentation will be less.

2 - Anticipative Long Term Forest Planning

Sophie D’Amours, Université Laval, Department of Mechanical

Engineering, Pavillon Adrien Pouliot, Quebec, QC, G1K 7P4,

Canada, [email protected], Mathieu Bouchard,

Martin Simard, AchilleBenjamin Laurent, Sebastien Lemieux

This presentation will describe the challenging link between long term forest planning and industry development. It will showcase web-based advanced planning tools (SylviLab and LogiLab) supporting group decision making.

Anticipation of the impact of the long term forest plan on industry and carbon balance will be explained. Multiple objectives are considered when selecting the best forest plans. Case studies from Canada will be presented.

3 - Preventing Orphan Stands in Spatially Explicit Forest

Management Planning

Marc McDill, Associate Professor, Pennsylvania State School of

Forest Resources, 310 Forest Resources Building, University Park,

PA, 16802, United States of America, [email protected],

Susete Marques, Jose G Borges

For a variety of reasons, in spatially-explicit harvest scheduling models some management units typically will not be harvested within the planning horizon.

Small unharvested management units may be too small to allow a viable harvesting operation, however, leaving these units as unusable “orphans” at the end of the planning horizon. We present a multi-objective formulation that minimizes the number of orphan units.

275

4 - Multiobjective Forest Management Model with Uncertain

Weights

Cristian D. Palma, Universidad del Desarrollo, Faculty of Industrial

Engineering, Santiago, Chile, [email protected],

John D. Nelson

Defining weights in a multiobjective optimization framework is a difficult task, especially when many actors are involved in the decision-making process. We present a robust version of a multiobjective multiperiod forest management problem in which the weights are defined as a continuous range of possible values, and show that robust decisions produced more stable outcomes through the planning horizon than deterministic ones when weights may change over time.

TA42

42- North 227 C- CC

Shared Mobility Systems

Sponsor: Transportation Science & Logistics/ Urban Transportation

Sponsored Session

Chair: Virot Chiraphadhanakul, PhD Candidate, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA, 02139,

United States of America, [email protected]

1 - A Transportation Market for Ridesharing

Maged Dessouky, Professor, University of Southern California,

University Park Campus, Los Angeles, CA, 90007,

United States of America, [email protected], H. Xu, M. Furuhata,

Fernando Ordonez, Sven Koenig, K. Daniel, X. Wang

We describe mechanisms and algorithms for negotiating routes and prices to facilitate a transportation market for ridesharing. We present a pricing model that fairly allocates the cost of the trip among all the passengers as well as ensures that it is always cheaper to request a ride as early as possible, an algorithm that routes and schedules the vehicles that minimizes the cost and passenger travel times, and a network equilibrium model that determines the optimal number of riders.

2 - Evaluating Impacts of Integrating Vehicle-Sharing Systems with an Existing Public Transit Network

Virot Chiraphadhanakul, PhD Candidate, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA, 02139,

United States of America, [email protected], Cynthia Barnhart

Using publicly available transit schedules, we construct a public transit graph that enables us to identify an optimal transit route from one transit node to another based on various objective functions and constraints. We then augment the graph with additional nodes and arcs corresponding to different designs of vehiclesharing systems and evaluate their impacts on trip times, commuter choices, and emissions. Empirical results from Boston metro area will be discussed.

3 - Managing Public Bike Sharing Systems by Demand Profile and

Temporary Manpower Allocation

I-Lin Wang, Associate Professor, Department of Industrial and

Information Management, National Cheng Kung University, No.1

University Rd., Tainan, Taiwan-ROC, [email protected]

We consider the surveyed OD demand profile as a fixed input that defines the bikes emanating from each station in a proportional fashion. To achieve better service quality without moving bikes among stations by bike-repositioning vehicles, our first model calculates where and when to put caretakers for receiving more bikes in over-capacitated stations, whereas our second model seeks the best amount of volunteer riders to be invited for riding between some OD links at specific periods.

4 - Comparing Optimal Relocations with Realistic Simulated

Relocation Policies in One-way Carsharing

Diana Jorge, University of Coimbra, Rua Luìs Reis Santos,

Coimbra, Portugal, dianarjorge[email protected], Cynthia Barnhart,

Gonçalo Correia

The objective of this paper is developing a mathematical model that optimizes vehicle relocation operations given a trip pattern for a particular set of stations. A simulation model is also developed to apply different relocation policies for the same instances. We conclude that it is not possible to reach the optimum results using relocation policies. However both approaches lead to a significant increase in the profit of the company.

TA43

INFORMS Phoenix – 2012

TA43

43- North 228 A- CC

Joint Session RAS/TSL: Railroad

Optimization Models

Sponsor: Railway Applications & Transportation Science & Logistics

Sponsored Session

Chair: Aihong Wen, Manager Operations Research, CSX, 500 Water

Street J300, Jacksonville, FL, 32224, United States of America, [email protected]

1 - Overview of Decision Support Systems for Intermodal

Operations

Yudi Pranoto, Norfolk Southern Corporation, 1200 Peachtree

Street N.E., Atlanta, GA, 30309, United States of America, [email protected], Hyun-suk Yoon, Clark Cheng

We describe the application of forecasting and optimization methods in an asset management tool in used by the Intermodal Operations at Norfolk Southern.

2 - Movement Planner

Srinivas Bollapragada, GE Global Research, 1 Research Circle,

Niskayuna, NY, 12309, United States of America, [email protected]

A movement plan is a detailed plan for moving trains in a rail network over a time horizon of 8 to 12 hours. It includes planning meet locations where trains traveling in opposite directions can get past each other as well as pass locations where fast trains can get past slower trains traveling in the same direction. We developed and implemented a novel algorithm at a major railroad company to automatically dispatch trains to increase average train velocities and save up to

$800 million a year.

3 - Improve Operations by Drayage Optimization

Xiaoqing Peter Sun, Sr. Engineer, Schneider National, 3101 S.

Packerland Dr., Green Bay, WI, 54313, United States of America, [email protected], Kendall Bailey

An optimization decision support tool has been developed at Schneider National to assist dispatching decisions. The tool utilizes advanced mathematical programming techniques such column generation, resource constrained shortest path problem to generate driver tours. The tool meets all practical constraints such as different appointment windows, freight types, rush hours, different driver types. The tool has been utilized for the last few years and significant cost savings have been achieved.

4 - Train Design Optimization in a Congested Railroad Network

Abdullah Al Khaled, PhD Student, Mississippi State University,

Department of Industrial and Systems Engg., McCain Engineering

Building, P.O. Box 9542, Mississippi State, MS, 39762, United

States of America, [email protected], Mingzhou Jin

Train design in railways consists of two tasks: block-to-train assignment and train routing. It seeks to identify optimal routes for trains and their associated blocks and is subject to various capacity and operational constraints at tracks and yards.

A mixed integer program is formulated for this problem considering nonlinear relationship between travel time and volume in a congested network. A two-stage heuristic algorithm is proposed to attack the computational burden for real-world networks.

2 - The Role of Power in the Medical Equipment Supply Chain

Jurriaan de Jong, The Ohio State University, 3238 Ainwick Rd,

Columbus, OH, 43221, United States of America, [email protected], W.C. Benton

Many health organizations are members of large powerful Group Purchasing

Organizations (GPOs). We empirically analyze the role of supply chain power and the presence of a middleman, a GPO, in the medical equipment supply chain. In particular we study the effects first on relationships between manufacturers and the health organizations and between GPOs and the health organization and second on the performance of the respective supply chain members.

3 - Inventory Positioning in Clinical Trial Supply Chains

Yao Zhao, Associate Professor, Rutgers Business School,

1 Washington Park, Newark, NJ, 07102, United States of America, [email protected], Adam Fleischhacker, Anh Ninh

One bottleneck in clinical trials is the slow patient recruitment. Companies typically address this issue by recruiting patients globally, which, however, resulted in significant overages of clinical supplies. In this study, we introduce a class of mathematic models to capture the unique attributes of clinical trial supply chains. We show how to optimally position inventory in such supply chains, and test the effectiveness of the model in a real-world example.

4 - Mitigating the Effects of Drug Shortages in the Global

Healthcare Supply Chains

Raja Jayaraman, Assistant Professor, Khalifa University of Science

Technology and Research, Department of Industrial & Systems

Engg., Al Saada St. (19th) & 2nd Street, Abu Dhabi, United Arab

Emirates, [email protected], Young-Ji Byon,

Young Seon Jeong

Globally healthcare supply chains face critical challenges in healthcare delivery due to shortages of essential and life saving drugs. It is often too difficult to distinguish between isolated and regular shortages. In this presentation we assess the information adequacy and resources needed to effectively manage drug shortages.

TA45

45- North 229 A- CC

Panel Discussion: How to Undertake

Collaborative Research

Sponsor: Junior Faculty Interest Group

Sponsored Session

Chair: Halit Uster, Industrial and Systems Engineering, Texas A&M

University, College Station, TX, United States of America, [email protected]

1 - Panel Discussion: How to Undertake Collaborative Research

Moderator: Halit Uster, Industrial and Systems Engineering, Texas

A&M University, College Station, TX, United States of America, [email protected], Panelists: Diego Klabjan, Eva Lee, Jon Lee,

Warren Powell

Timely with increasing attention and excitement about collaborative and interdisciplinary research, the panel will discuss the making and the sustainment of successful collaborations among academics as well as between academics and practitioners.

TA44

44- North 228 B- CC

Supply Chain: Issues in Healthcare

Contributed Session

Chair: Raja Jayaraman, Assistant Professor, Khalifa University of

Science Technology and Research, Department of Industrial & Systems

Engg., Al Saada St. (19th) & 2nd Street, Abu Dhabi, United Arab

Emirates, [email protected]

1 - Pulling, Pooling, and Contracting under Heterogeneous

Consumers

Gerardo Pelayo, Zaragoza Logistics Center, C/ Bari 55, Edificio

Nayade 5 (PLAZA), Zaragoza, Spain, [email protected],

Mustafa Cagri Gurbuz

We model a pharmaceutical manufacturer who owns a drug that a applies to a single (stochastic size) category of patients and can exert innovation effort aimed at creating a second (deterministic size) category who benefits differently from the drug. We analyze the stocking and effort decisions when reservation is not allowed under dedicated versus a single sales channel and find conditions under which pooling does not reduce total inventory nor achieves the efficient level of effort.

TA46

46- North 229 B- CC

Disaster and Emergency Response I

Contributed Session

Chair: Sandro Paz, PhD Candidate / Associate Professor, University of

South Florida / Pontificia Universidad Catolica del Peru, 18125

Birdwater Dr, Tampa, FL, 33647, United States of America, [email protected]

1 - Dispatching and Relocation of Heterogeneous

Emergency Vehicles

Elham Sharifi, PhD Candidate, University of Maryland, 1179

Glenn Martin Hall, College Park, United States of America, [email protected], Ali Haghani

Emergency response services play crucial roles in all communities and can minimize the adverse effects of emergency incidents by decreasing response time.

Response time is not only related to the dispatching system, but also has a close relationship to the coverage of the whole network by emergency vehicles. The goal of this research is to develop a model that will dynamically dispatch some vehicles to emergency sites and also relocate others to provide better coverage for the future demands.

276

INFORMS Phoenix – 2012

2 - Critical Infrastructure Management for

Telecommunication Networks

Haibo Wang, Associate Professor, Texas A&M International

University, 5201 University Blvd, Laredo, TX, 78041, United States of America, [email protected], Bahram Alidaee, Wei Wang

Telecommunication network infrastructures play an important role to maintain the stability of society worldwide. This paper presents a critical infrastructure detection model to discover the interdependency based on the model from social network and new communication pathways. The policy and procedure of protecting critical infrastructures are discussed and computational results from the proposed model are presented.

3 - Optimize Pre-disaster Investment to Strengthen a Vulnerable

Transportation Network

Kwangho Kim, Department of Civil and Environmental

Engineering, University of California, 416F McLaughlin Hall,

Berkeley, United States of America, [email protected],

Yoonjin Yoon

A bi-level programming algorithm is developed to optimize pre-disaster resource allocation to strengthen a vulnerable transportation network by considering system-wide network vulnerability. The proposed algorithm is tested on a medium-size network to evaluate the benefit of employing vulnerability analyses to suitably predefine choice sets of candidate investment strategies. The proposed methodology would advance the existing knowledge on pre-disaster infrastructure management.

4 - A Framework to Evaluate System-wide Vulnerability of a

Transportation Network under Disasters Events

Sehyun Tak, PhD. Candidate, KAIST, Yuseong-Gu, 335

Gwahangno, Daejeon, 305-701, Korea, Republic of, [email protected], Kwangho Kim, Yoonjin Yoon

It is important to assess vulnerability of transportation network in disaster management. In our study, we identify critical links of transportation network under user equilibrium by evaluating system travel time in the pre and post disasters. We adopt a simulation study and apply a set of scenarios by varying demand as well as network connectivity.

5 - Antiviral Resistance and Mitigation of Pandemic Influenza

Sandro Paz, PhD Candidate/Associate Professor, University of

South Florida/Pontificia Universidad Catolica del Peru, 18125

Birdwater Dr, Tampa, FL, 33647, United States of America, [email protected], Alex Savachkin

We are building a large-scale simulation optimization framework for antiviral based mitigation of pandemic influenza. The model considers an oseltamivirsensitive strain and a resistant strain with low/high fitness cost. Mitigation strategies include treatment of symptomatic cases and chemoprophylaxis of preand post-exposure cases.

TA47

47- North 230- CC

Dynamic Traffic Assignment I - New

Modeling Approaches

Sponsor: Transportation Science & Logistics/ Intelligent

Transportation Systems (ITS)

Sponsored Session

Chair: Chung-Cheng Lu, National Taipei University of Technology,

1 Section 3, Chung-Hsiao East Road, Taipei, Taiwan - ROC, [email protected]

1 - A Game-Theoretic Perspective on Dynamic Network

Equilibrium

Stephen Boyles, Assistant Professor, The University of Texas,

Austin, 1 University Station C1761, Austin, TX, United States of

America, [email protected], Christopher Melson

Dynamic network equilibrium is much more complex than static equilibrium.

Equilibrium may not be unique, and may not even exist in pure strategies, which is significant for simulation-based dynamic traffic assignment. This presentation discusses several small examples of these issues, and introduces a game-theoretic perspective which may be useful in resolving them.

2 - A Nonlinear Integer Program for Vehicle-based Dynamic

Traffic Assignment

Chung-Cheng Lu, National Taipei University of Technology,

1 Section 3, Chung-Hsiao East Road, Taipei, Taiwan-ROC, [email protected], Kuilin Zhang, Xuesong Zhou

We propose an innovative nonlinear integer program for vehicle-based dynamic traffic assignment problems. The objective function is a vehicle-based gap function that measures the deviation from dynamic user equilibrium. This model describes dynamic traffic propagations and handles first-in-first-out requirements

TA48

with a set of linear constraints. In addition, a set of time connectivity constraints provides facets of the convex hull, so the solutions of the LP relaxation of the integer program are almost always integral, and rounding heuristics or branchand-bound are not required. A numerical example is presented to demonstrate the integral property of the model.

3 - Price of Anarchy in Dynamic Traffic Networks:

A Computational Study

Satish Ukkusuri, Associate Professor, Purdue University,

550 Stadium Mall Drive, G175F, West Lafayette, IN, 47906,

United States of America, [email protected], Kien Doan

Price of Anarchy is described as the worst case difference between the equilibrium and optimum solution thereby measuring the inefficiency in traffic networks. This study will demonstrate the price of anarchy in dynamic traffic networks. A brief review of the dynamic equilibrium and dynamic system optimum models will be provided. Extensive computational analysis will demonstrate the price of anarchy in various traffic networks. Open questions of interest to the research community will be discussed.

4 - Stability and Convergence in Large-scale Dynamic

Traffic Assignment

Stephen Boyles, Assistant Professor, The University of Texas,

Austin, 1 University Station C1761, Austin, TX, United States of

America, [email protected], C. Matthew Pool

This work reviews, implements and compares existing techniques for finding equilibrium solutions in simulation-based dynamic traffic assignment and proposes novel approaches. Numerical experiments assess the convergence and stability properties of these approaches in realistic-sized networks.

TA48

48- North 231 A- CC

Port Operations and Shipping Logistics

Sponsor: Transportation Science & Logistics/ Freight

Transportation & Logistics

Sponsored Session

Chair: Yun Fong LimAssistant Professor, Singapore Management

University, 50 Stamford Road, #04-01, Singapore, SG, Singapore, [email protected]

1 - A Quay Crane System That Self-Recovers from Random Shocks

Yun Fong Lim, Assistant Professor, Singapore Management

University, 50 Stamford Road, #04-01, Singapore, SG, Singapore, [email protected], Yan Zhang

The main challenge for a container terminal is to maximize its throughput using limited resources subject to various operational constraints. Traditional approaches often require re-planning when the system is subject to random shocks. We propose an operating protocol to coordinate the cranes so that the system spontaneously recovers from shocks and persistently restores its throughput near its full capacity.

2 - Optimal Scheduling for the Inland Transport of Containers:

Facing the Trade-off of Costs, Efficiency and Sustainability

Stefano Fazi, Eindhoven University of Technology, Lismortel, P.O.

Box 513, Eindhoven, 5600 MB, Netherlands, [email protected],

Jan Fransoo, Tom van Woensel

We present an allocation model for a particular container supply chain where multiple sea terminal quays and a hinterland terminal are considered. From the moment a container is dropped off from the vessels at the sea terminals, several options for the final transportation to the hinterland can be considered: trucks, barges and trains. The logistic planner decides the best allocation for the containers to the available means of transport. The objective is to minimize the costs for the transportation while guaranteeing a high service level and low congestion at the sea terminal.

3 - Berth Allocation Problem Considering Fuel Consumption and

Vessel Emissions in a Decentralized Decision Environment

Yueran Zhuo, University of Massachusetts, Isenberg School of

Management, Amherst, MA, 01003, United States of America, [email protected], Yuquan Du

We focus on the berth allocation problem considering fuel consumption and vessel emissions in a decentralized decision environment. We model and solve this problem by deploying a multi-agent system, where the optimal sailing speed of each vessel is obtained by an iterative negotiation process between the terminal and vessels. Experimental results show that the model can significantly reduce fuel consumption and vessel emissions in sailing periods.

277

TA49

INFORMS Phoenix – 2012

4 - A Proactive International Shipment Consolidation Strategy

Wooseung Jang, Professor, University of Missouri, E3437 Lafferre

Hall, Columbia, MO, 65211, United States of America, [email protected], Na Deng

We consider an integrated multi-commodity consolidation and shipping problem, which occurs in an international supply network. We propose a proactive strategy that consolidates shipment, considering both ocean and subsequent inland transportation activities. A combinatorial optimization model is proposed and numerical analysis shows significant cost savings of our approach.

TA49

49- North 231 B- CC

Joint Session TSL/SPPSN: Evacuation Planning

Sponsor: Transportation Science & Logistics & Public Programs,

Service and Needs

Sponsored Session

Chair: Yue Liu, University of Wisconsin-Milwaukee, P.O. Box 784,

Milwaukee, WI, 53201, United States of America, [email protected]

1 - Integrated Demand-supply Management Framework and

Risk-based Routing for Evacuation Operations

Yu-Ting Hsu, Purdue University, 3000 Kent Ave., Nextrans Center,

West Lafayette, IN, 47906, United States of America, [email protected], Srinivas Peeta

We propose an operational framework for mass evacuation which integrates demand and supply management perspectives. It incorporates evacuee behavior models for evacuation recommendation and route guidance. A risk-based routing strategy is adopted that seeks to evacuate the population with the highest risk first by disseminating information to it. It introduces a risk management perspective and also circumvents expensive iterative computation, thereby enabling real-time evacuation operations.

2 - Optimal Decision-making between Signalized and

Uninterrupted Flow Strategies during Evacuation

Zhenke Luo, University of Wisconsin-Milwaukee, P.O. Box 784,

Milwaukee, WI, 53201, United States of America, [email protected], Yue Liu, Jing Mao

This paper presents a bi-level network optimization model to determine the optimal set of intersections in the evacuation network for implementing uninterrupted flow and signal control strategies, respectively, which can yield the maximum evacuation operational efficiency and the best use of available budgets.

The proposed model is solved by a GA-based heuristic.

3 - Determination of the Onset and Duration of Contraflow

Operations under Uncertainties

Yingyan Lou, Assistant Professor, The University of Alabama, 260

H.M. Comer, Tuscaloosa, AL, 35487, United States of America, [email protected], Daniel Fonseca, Gary Moynihan, Peiheng Li,

Saravanan Gurupackiam

This research addresses the determination of the onset and duration of contraflow operations for hurricane evacuation planning. A modeling framework is developed featuring (1) an underlying traffic flow model to describe traffic dynamics; (2) incident identification and characterization using artificial intelligence; (3) various optimization techniques for decision making under uncertainties; and (4) a computer-based decision support system. A case study of

I-65 in Alabama is provided.

4 - Integrated Modeling of Hurricane Evacuation Decision Making:

An Interdisciplinary Perspective

Binh Luong, Purdue University, 550 Stadium Mall Drive, G175F,

West Lafayette, IN, 47906, United States of America, [email protected], Samiul Hasan, Kien Doan,

Rodrigo Mesa-Arango, Sattish Ulkusuri

The goal of this talk is to describe an integrated modeling approach for hurricane evacuation modeling from behavioral and simulation modeling. Various behavioral dimensions that influence households to evacuate will be discussed including the timing of evacuation, the destination choice and route choice. The models are estimated with data from Hurricane Ivan. A large scale simulation model will be demonstrated to show the efficiency of the evacuation clearance times.

TA50

50- North 231 C- CC

Military Manpower Planning Models and Applications

Sponsor: Military Applications

Sponsored Session

Chair: Patrick DownesPhD Candidate, University of Texas/

United States Army, 2716 Dupree Lane, Austin, TX, 78748,

United States of America, [email protected]

1 - Manpower Planning Models for U.S. Army Medical Department:

Optimization and Simulation Approaches

Patrick Downes, PhD Candidate, University of Texas/

United States Army, 2716 Dupree Lane, Austin, TX, 78748,

United States of America, [email protected]

The U.S. Army uses optimization models to manage manpower policies. We present how current objective force models of the Army Medical Department can be extended to provide optimal manpower policies for the current force. We study two models: a linear goal program and a system dynamics model. The first is an extension of current models used today and the system dynamics model includes the study of how an individual’s probability of continuation is influenced by additional, important factors.

2 - Project Selection and Manpower Planning in the Colombian

Navy

Andrès L. Medaglia, Associate Professor, Universidad de los Andes,

Departamento de Ingenieria Industrial, Carrera 1 Este # 19 A-40,

Bogotá, Colombia, [email protected], Miguel Barrios,

Wilson Flórez, Maria A. Londoño, Jorge Sefair, Juan D. Palacio,

Juan D. Ortiz, Flor Sánchez, Mauricio Mejìa

To make a better use of its resources and to leverage negotiation with the Defense

Ministry and the Colombian Planning Department, the Colombian Navy (ARC, for its acronym in Spanish) has developed a Web-based decision support system to select and schedule a bank of projects. Many of these projects (e.g., acquisition of a new coast guard patrol boat) have a direct impact on manpower planning.

Thus, a second decision support tool allows the ARC to estimate costs and properly plan its workforce.

TA51

51- North 232 A- CC

Tutorial: Managing Spare Parts

Sponsor: Military Applications

Sponsored Session

Chair: Thomas Willemain, Vice President, Smart Software, Inc.,

4 Hill Road, Belmont, MA, 02478, United States of America, [email protected]

1 - Tutorial: Managing Spare Parts

Thomas Willemain, Vice President, Smart Software, Inc.,

4 Hill Road, Belmont, MA, 02478, United States of America, [email protected]

This tutorial is for people new to this topic, covering the basics and emphasizing applied topics: key issues in management of military (and public sector) spare parts, performance metrics, probability models used in forecasting demand, software systems, connections to basic inventory control theory, and pointers to the literature.

278

INFORMS Phoenix – 2012

TA52

52- North 232 B- CC

Evaluating the Controllability of a System of Plug-in

Hybrid Electric Vehicle Charging Stations

Sponsor: Energy, Natural Res & the Environment/Energy

Sponsored Session

Chair: Victoria Chen, University of Texas at Arlington, Department Ind.

& Manuf. Sys. Engr., Arlington, TX, United States of America, [email protected]

1 - Adaptive Design for Controllability of Charging Stations for

Plug-in Hybrid Electric Vehicles

Amirhossein Khosrojerdi, Graduate Research Assistant, University of Oklahoma, 202 W. Boyd Street, Norman, OK, 73071, United

States of America, [email protected], Piampoom Sarikprueck,

Asama Kulvanitchaiyanunt, Janet K Allen, Victoria Chen,

Minting Xiao, Jay Rosenberger, Wei-Jen Lee, Salman Ahmed,

Farrokh Mistree

A system of PHEV charging stations is evolving as parameters change over time.

The simultaneous design of the control systems and the infrastructure to support it will result in both enhanced controllability and provide a more efficient design process. We focus on the modular view of designing a system whose performance is determined by their ability to be controlled and on a method which produces robust systems which can evolve effectively. The focus is on the method, rather than the results.

2 - Modeling and Forecasting Towards Adaptive Design of a

System of PHEV Charging Stations

Piampoom Sarikprueck, The University of Texas at Arlington,

416 Yates Street, Nedderman Hall Room 517-518,

Arlington, TX, 76011, United States of America, [email protected], Wei-Jen Lee,

Asama Kulvanitchaiyanunt, Victoria Chen, Jay Rosenberger

The proposed model for PHEV charging stations consists of energy resources, controller, battery, converter, and inverter. The main energy resources are utility grid integrated with wind from a single wind farm and solar energy generated at each station. The charging stations will provide DC fast charging and can sell back energy to the grid. Forecasting results of wind and solar energy as well as market prices are carried out to provide information for system design and optimal dynamic control.

3 - Discrete Event Simulation of Demand at PHEV Charging

Stations

Minting Xiao, University of Oklahoma, 202 W. Boyd St.,

Room 116, Norman, OK, 73019-0631, United States of America, [email protected], Salman Ahmed, Janet K Allen, Farrokh Mistree

Discrete-event simulation is used to predict the demand for energy at a PHEV charging station. Ranges of demand are predicted based on historical projections and anticipated policy decisions and this information is used in the discrete event simulation. Further uncertainty is incorporated in the model based on expected fluctuations in charging at different times of day. The simulation of demand from one charging station can be then extended to the system of charging stations.

4 - Stochastic Dynamic Control for PHEV Charging Stations

Asama Kulvanitchaiyanunt, The University of Texas at Arlington,

416 Yates Street, Arlington, TX, United States of America, [email protected], Piampoom Sarikprueck,

Victoria Chen, Jay Rosenberger, Wei-Jen Lee

An infinite horizon policy for a continuous-state stochastic dynamic program is proposed to control a system of plug-in hybrid electric vehicle (PHEV) charging stations. The martingale model of forecast evolution is used to represent uncertainties for wind farm power generation, solar power generation at each charging station, and nodal market price. Due to the large-scale nature of the problem, an approximate dynamic programming based on design and analysis of computer experiments is developed.

TA53

53- North 232 C- CC

Advanced Models and Algorithms for Power Grid

Design and Operations

Sponsor: Energy, Natural Res & the Environment/Energy

Sponsored Session

Chair: Bo Zeng, Assistant Professor, University of South Florida, 4202 E

Fowler Ave ENB118, Tampa, FL, 33620, United States of America, [email protected]

1 - Fully Decentralized AC Optimal Power Flow Algorithms

Andy Sun, Postdoctoral Researcher, IBM Thomas J Watson

Research Center, 1101 Kitchawan Road, Yorktown Heights, NY,

10598, United States of America, [email protected], Dzung

Phan, Soumyadip Ghosh

Motivated by the increasing complexity in the control of electric power systems especially in a smart grid environment, we propose fully decentralized algorithms to solve AC OPF problems. The key feature of our algorithms is a complete decentralization of computation down to nodal level, where the problem is solved by individual nodes with only local knowledge. Our algorithms have computation complexity of linear growth and can adapt to network topology changes without central coordination.

2 - Worst-case Contingency-constrained Expansion Planning for

Power Systems

Neng Fan, Sandia National Laboratories, P.O. Box 5800, MS 1326,

Albuquerque, NM, 87185, United States of America, [email protected]

In this talk, we include the constraints of the N-k contingency into the planning process, such that the system can still run normally in case of contingencies with multiple simultaneous loss of the system elements. Robust optimization method is used to identify the worst-case contingency for analysis. In the operation level, the technique of line switching is added to check the benefits comparing with the situation without line switching.

3 - Robust Temperature Modulated Unit Commitment Problem

Bo Zeng, Assistant Professor, University of South Florida, 4202 E

Fowler Ave., ENB118, Tampa, FL, 33620, United States of

America, [email protected], Anna Danandeh

The efficiency of gas generators is heavily affected by the temperature of the inlet air, which could be 10% loss. Hence, classical unit commitment models may not yield feasible solutions to meet load demand. We consider a two-stage unit commitment problem where temperature in next 24 hour is random, and present solution algorithms for this new type robust optimization problem.

4 - Two-stage Stochastic Unit Commitment Models With Energy

Storage and Demand Response

Yuping Huang, Graduate Research Assistant, West Virginia

University, 478 Harding Ave., Apt 7, Morgantown, WV, 26505,

United States of America, [email protected], Qipeng Zheng

Demand response is a key component of smart grid and energy storage is crucial for making variable renewable energy sources once they’re connected to power grid. This study is to develop two-stage stochastic programming models for unit commitment with energy storage, with demand response, and with both resources. Numerical experiments are conducted to compare the effects of both resources on power generation.

TA54

54- Regency Ballroom A- Hyatt

Business Innovation Cases in Korean Firms

Cluster: KINFORMS (Korea Chapter—INFORMS)

Invited Session

TA54

Chair: Sang Hyung Ahn, Seoul National University, School of Business,

Seoul, Korea, Republic of, [email protected]

1 - Process-based Service Quality Assessment of

Automobile Maintenance Service

Hosun Rhim, Professor, Korea University Business School,

Anam-dong, Seongbuk-gu, Seoul, 136-701, Korea, Republic of, [email protected], Hye-Jun Gil, Geon-Ha Kim, Kil-Hang Wang

We measure and evaluate the service quality of auto repair service. Service is measured by stages of service process. We investigate relationships among the service quality of auto repair center, customer loyalty toward auto maintenance center, and repurchase intention of automobile. We conducted an online survey and collected a sample of size 3,028.

279

TA56

2 - An Analysis of Rotating Savings and Credit Associations

Deung-Geon Ahn, KAIST, Industrial and Systems Eng.,

Daejeon, Korea, Republic of, [email protected],

Wanmo Kang, Kyoung-Kuk Kim, Hayong Shin

Rotating Savings and Credit Association (ROSCA) is a well-known and widely used informal financial association in many countries around the world.

However, the ROSCA has not been much investigated at the quantitative level.

This work analyzes the structure and probability, compared to bank transactions, of one type of ROSCAs which has a very long history in the East Asia. Unlike the previous studies, we approach the ROSCA using replication as often done in financial engineering. We identify some conditions that lead to existence of such private credit associations and optimal orders of participants in such systems.

3 - Supply Chain Risk Management Strategy

Chang Won Lee, Professor and Chair of Operations and Service

Management, Hanyang University, Seoul, 133-791, Korea,

Republic of, [email protected], Sang Hyung Ahn

The study present current supply chain risk management strategy in terms of global perspective. The background has been reviewed and research model is developed. Empirical study is conducted and results are analyzed and discussed.

The managerial insights for global supply chain risk management strategy has been provided.

4 - A Parametric Analysis for a Risk-averse Newsvendor

Sungyong Choi, Assistant Professor, Yonsei University,

1 Yonseidae-gil, Wonju, 220-710, Korea, Republic of, [email protected]

I consider a multi-product risk-averse newsvendor with the mean-CVaR

(Conditional Value-at-Risk) model. I establish a parametric analysis for the model.

Then, I prove the impacts of number of products and degree of risk aversion to the problem in independent demand case. For dependent demands, I conduct a special case analysis to obtain a closed-form solution. A numerical study confirms my analytical results to show how risk aversion and dependent demands jointly affect to the optimal solution.

5 - New Quality as a Competitive Weapon in Korean

Automotive Industry

Youn Sung Kim, Inha University, Korea, Republic of, [email protected]

Quality is one of the most important issues in many companies. After the Toyota crisis, many companies are trying to enhance the quality level and reduce the quality problem. So in this research, we try to find out “how to reenergize their quality focus”, especially in Korean automotive industry.

TA56

56- Curtis A- Hyatt

Joint Session TMS/ORG: Managing Product

Configuration and Improvement

Sponsor: Technology Management & Organization Science

Sponsored Session

Chair: Zhijian Cui, Assistant Professor of Operations Management, IE

Business School, Calle de Maria de Molina, 12, Madrid, 28006, Spain,

[email protected]

1 - Product Configuration, Ambidexterity and Firm Performance

Fabrizo Salvador, Professor, IE Business School,

Calle Maria de Molina 12-5, Madrid, Ma, 28006, Spain,

[email protected]

We identify two firm-level abilities–namely, product configuration effectiveness

(PCE) and product configuration intelligence (PCI)–that explain how product configuration impacts firm performance. We link PCE and PCI to exploration and exploitation and, following the ambidexterity argument, propose that their interaction is associated to superior operational and financial performance. We find empirical support the ambidexterity argument relative to sales, but not to profit.

2 - Joint Product Improvement by Client and Customer Support

Center: The Role of Gain-Share Contracts

Sameer Hasija, INSEAD, Fontainbleau, France, [email protected], Shantanu Bhattacharya, Alok Gupta

We study the role of different contract types in coordinating the joint product improvement effort of a client and a customer support center. The customer support center’s costly efforts include transcribing and analyzing customer feedback, analyzing market trends, and investing in product design. We show the role of gain-share contracts in resolving the agency issues present is such settings.

INFORMS Phoenix – 2012

3 - Optimizing the Resource Allocation of an Entrepreneur:

The Role of Learning Externalities

Onesun Yoo, University College London, Gower Street, London,

WC1E 6BT, United Kingdom, [email protected],

Bilal Gokpinar, Yufei Huang

We study the problem of an entrepreneur trying to sell his service to two potential buyers by providing them information about the service. Buyers learn and update their beliefs about the service quality both directly from the entrepreneur himself, and indirectly through the purchase decision of the other buyer. By developing an analytical model, we characterize this learning and purchasing process, and analyze the optimal resource allocation policy for the entrepreneur.

4 - Collaborative Search

Fabian Sting, Rotterdam School of Management, Erasmus

University, Burgemeester Oudlaan 50, Rotterdam, Netherlands,

[email protected], Jurgen Mihm, Christoph Loch

Search has become a widely accepted paradigm to describe innovation activities.

Formal models of search have incorporated a broad spectrum of different aspects relevant to search, such as cognition or organizational embedding. We contribute to the understanding of search by studying under which circumstances it is beneficial to have several organizational players search collaboratively.

TA57

57- Curtis B- Hyatt

Applied Probability and Financial Engineering

Sponsor: Applied Probability

Sponsored Session

Chair: Martin Haugh, Columbia University IE&OR, 500 West 120th

Street, Mudd 332, New York, NY, 10027, United States of America, [email protected]

1 - Transform Formulae for Linear Functionals of Affine Processes and their Bridges on PSD

Wanmo Kang, Assistant Professor, KAIST, 291 Daehak-ro,

Yuseong-gu, Daejeon, Korea, Republic of, [email protected],

Chulmin Kang

We derive transform formulae for linear functionals of affine processes and their bridges whose state space is a set of positive semidefinite matrices. We investigate the relationship between such transforms and certain integral equations. Our findings extend and unify the existing results on affine processes and squared

Bessel functionals, respectively. We also derive analytic expressions for Laplace transforms of some functionals of Wishart bridges. This is a joint work with

Chulmin Kang.

2 - Simulated Likelihood Inference for Jump-diffusions

Gustavo Schwenkler, Stanford University, 1045 Alma St., Palo

Alto, CA, 94301, United States of America, [email protected], Kay Giesecke

We develop likelihood estimators for the parameters of a discretely observed jump-diffusion with state-depend drift, diffusion and jump intensity functions.

The estimators are based on a novel representation of the transition density that can be computed exactly using an exact sampling algorithm. We provide conditions for consistency and asymptotic normality of the estimators. Numerical results illustrate our approach.

3 - Value and Policy Iterations with Information Relaxation to

Stochastic Dynamic Programming

Nan Chen, Assistant Professor, The Chinese University of Hong

Kong, Shatin, N.T., Hong Kong, Hong Kong-PRC, [email protected], Wei Yu

We use the information relaxation technique to develop a value-and-policy iterative method to solve discrete time dynamic programming problems. Both value and policy converges to the optimal ones within finite steps of iterations.

280

INFORMS Phoenix – 2012

TA58

58- Phoenix East- Hyatt

Advances in Stochastic Optimal Control and Learning II

Sponsor: Applied Probability

Sponsored Session

Chair: Ciamac Moallemi, Associate Professor, Columbia Business

School, 3022 Broadway, Uris 416, New York, NY, 10025,

United States of America, [email protected]

Co-Chair: Vivek Farias, Massachusetts Institute of Technology,

77 Massachusetts Avenue, Cambridge, MA, United States of America, [email protected]

1 - Feature Search on the Grassmanian Manifold for

Reinforcement Learning

Vivek Borkar, Indian Institute of Science, Deaprtment of Electrical

Engineering, Bangalore, 400076, India, [email protected],

K.J. Prabuchandran, Shalabh Bhatnagar

We address the problem of finding the best features in reinforcement learning with linear function approximation. Specifically, we consider TD(0) with linear function approximation for the infinite horizon discounted cost Markov decision process, wherein the number of `basis’ features is fixed. The feature vectors are adapted on a slower time scale by using a gradient descent on the Grassmanian manifold. This involves explicit gradient computation based on the

Edelman{Arias{Smith formula and a secondary learning algorithm based on the same. Convergence properties of the combined scheme are established using two time scale stochastic approximation theory and the standard `limiting o.d.e.’ approach to stochastic approximation. Simulation experiments show very good results.

2 - Means, Variances and Limiting Distributions in Finite-horizon

Markov Decision Problems

Alessandro Arlotto, Duke University, The Fuqua School of

Business, 100 Fuqua Drive, Durham, NC, 27708, United States of

America, [email protected], J. Michael Steele,

Noah Gans

The literature for Markov Decision Problems is huge but, despite a shared culture with the rest of applied probability, there are surprisingly few limit theorems for general classes of MDPs. This talk focuses on martingale (and other) methods that work well to obtain properties of the variance, and possibly of the distribution, for the total reward of some particular classes of finite-horizon MDPs.

3 - Information Relaxation Bounds for Infinite Horizon

Dynamic Programs

David Brown, Duke University, 100 Fuqua Drive, Durham, NC,

United States of America, [email protected], Martin Haugh

We study the use of information relaxations to compute upper bounds for a general class of infinite horizon stochastic dynamic programs. A challenge in using this approach for such problems is that probabilities in general depend on actions: even with perfect information, the subproblems may not be easy to solve.

We discuss some approaches for dealing with this challenge and illustrate these approaches in some applications.

TA59

59- Phoenix West- Hyatt

Optimal Control Policies I

Sponsor: Applied Probability

Sponsored Session

Chair: Ananth Krishnamurthy, University of Wisconsin-Madison, 1513

University Avenue, ME 3258, Madison, WI, United States of America, [email protected]

1 - Dynamic Inventory Control with Limited Capital and

Short-term Financing

Xiting Gong, University of Michigan, 1205 Beal Avenue,

Room 1815, Ann Arbor, MI, 48109, United States of America, [email protected], Xiuli Chao, David Simchi-Levi

We study a dynamic inventory control problem where a firm with some limited initial capital periodically purchases a product from a supplier and sells it to a market with random demands. In each period, besides using its own capital, the firm can also borrow a short-term loan to purchase the product, with the loan interest being an increasing convex function of the loan size. We characterize the structure of the optimal inventory policy that maximizes the firm’s expected terminal wealth.

TA60

2 - Throughput Maximization for Tandem Queues with

Synergistic Servers

Xinchang Wang, Georgia Institute of Technology, Room 316 Main

Building, 755 Ferst Drive, NW, Atlanta, GA, 30332, United States of America, [email protected], Sigrún Andradóttir,

Hayriye Ayhan

We consider tandem queues with servers that work more efficiently in teams than on their own. The synergy among collaborating servers can be station-dependent.

We provide a complete characterization of the optimal policy for Markovian systems with two stations and two servers. For larger systems, we provide sufficient conditions that guarantee that the optimal policy has all servers working together at all times.

3 - Optimal Production Inventory Strategies for Manufacturing

Systems with Seasonal Demands

Sanket Bhat, University of Wisconsin-Madison, 1513 University

Avenue, Room 3251, Madison, WI, 53726, United States of

America, [email protected], Ananth Krishnamurthy

We analyze manufacturing firms that can vary production capacities and target inventory levels to adapt to the seasonal demands for products. Based on a dynamic programming model, we identify and characterize the structure of the optimal policies. We also identify the optimal production capacities and length of the production runs for each season.

4 - Shortest Queue and Polling

Ivo Adan, Eindhoven University of Technology, Den Dolech 2,

Eindhoven, Netherlands, [email protected], Vidyadhar Kulkarni

In this talk we consider a two-station cyclic polling system with exhaustive service. The service times are exponential. Customers arrive to this system according to a Poisson stream and join the shortest queue (including the customer in service, if any). When both queues are equal they join the station where the server is. For this system we derive the steady-state joint queue length distribution.

TA60

60- Remington- Hyatt

Airport Capacity Frontier Analysis: Current Methods,

Tools and Findings: Part I

Sponsor: Aviation Applications

Sponsored Session

Chair: Dipasis Bhadram, Senior Economist, USDOT/FAA, 800

Independence Avenue, Washington, DC, DC, 20009,

United States of America, [email protected]

1 - Using Runway Assignment to Help Assess Aircraft Taxi-in Delay

Derek Robinson, Operations Research Analyst, Federal Aviation

Administration, 800 Independence Avenue SW, Washington, DC,

20591, United States of America, [email protected]

The objective of this study is to investigate aircraft taxi-in times at Core 30 airports. Most taxi-in analyses compare actual taxi times to a nominal time which only considers the airport and the operating carrier. We include runway assignment in our analysis. Using runway ends to determine the nominal taxi time can change the conclusions drawn from an assessment of those taxi times. In this investigation, we plan to compare delay results produced using both methods of defining a nominal time.

2 - The 2012 Update to the Airport Capacity Benchmarks

Jennifer Gentry, Lead Analyst, CAASD/MITRE Corporation,

7515 Colshire Avenue, McLean, VA, 22102-7539,

United States of America, [email protected]

The FAA has updated its Airport Capacity Benchmark Report. The report, first published in 2001 and updated in 2004, presents benchmark capacity ranges derived from Facility Reported Rates and modeled capacities. The 2012 update was motivated by changes in aviation, construction of new runways, and improved modeling techniques.

3 - runwaySimulator

Peter Kuzminski, Principal Staff, CAASD/MITRE Corporation,

7515 Colshire Avenue, McLean, VA, 22101,

United States of America, [email protected]

Come learn about The MITRE Corporation’s runwaySimulator simulation tool, used to estimate airport capacities for the Federal Aviation Administration’s

Airport Capacity Benchmark Report, the Future Airport Capacity Task, systemwide analyses, and other domestic and international work. An improved model and software will soon be available to the public.

281

TA61

INFORMS Phoenix – 2012

TA61

61- Russell- Hyatt

Homeland Security Analytics - I

Cluster: Applications in Emergency Management and

Terrorism Security

Invited Session

Chair: Barry EzellChief Scientist, Virginia Modeling, Analysis and

Simulation Center, 1030 University Blvd, Suffolk, VA, 23435,

United States of America, [email protected]

1 - Challenges for Public Sector Analytics

Evan Levine, Chief Scientist, DHS Risk and Capability Analytics,

Risk and Decision Analysis Directorate, 201 Varick St., 5th Flr-

NUSTL, New York, NY, 10014, United States of America, [email protected]

The growth of analytics in the public sector faces different challenges than in the private sector. Though opportunities to apply analytics in the public sector abound, cultural, organizational and technical challenges must be surmounted before government agencies can claim to be fully developed, enterprise-wide, analytically competitive organizations. The discussion will include examples of these challenges, as well as the progress being made to surmount them.

2 - Using Computable General Equilibrium Models to Estimate the

Economic Impacts of Homeland Security Hazards

Tony Cheesebrough, Director, DHS-Risk and Capability Analytics,

61 Walnut Avenue, Takoma Park, MD, 20912, United States of

America, [email protected], Antonio Kirson,

Jessica Smith, Nong Nai, Ryan Wise

Computable General Equilibrium (CGE) Models represent the state of the art in economic impact analysis. With assistance from the University of Southern

California’s Center for Risk and Economic Analysis of Terrorism Events

(CREATE), the U.S. Department of Homeland Security (DHS) has recently begun to develop a CGE model to advance the input-output (I-O) models currently used to estimate losses associated with terrorism, natural hazards, transnational crimes, and industrial accidents.

3 - Designing Independent Verification and Validation for DHS

Terrorism Risk Analysis

Barry Ezell, Chief Scientist, Virginia Modeling, Analysis and

Simulation Center, 1030 University Blvd, Suffolk, VA, 23435,

United States of America, [email protected], Dennis Buede,

Joe Tatman, John Lathrop

This presentation presents the research and design for conducting independent verification and validation of terrorism risk analysis. The research synthesizes generally accepted best practices in U.S. Department of Defense and leading scholarly research. The design was then applied to a specific risk assessment. The presentation describes many challenges and lessons learned in the process.

TA62

62- Borein A- Hyatt

Bidding in Auctions

Cluster: Auctions

Invited Session

Chair: Lin Hao, Assistant Professor, University of Notre Dame, Mendoza

College of Business, Notre Dame, IN, 46556, United States of America, [email protected]

1 - Strategic Bidding in Reverse Split-award Auctions

Kemal Guler, Hewlett Packard Labs, 1501 Page Mill, Palo Alto, CA,

United States of America, [email protected], Xiaoyong Zheng,

Stefan Mayer, Martin Bichler

We develop two competing first-price sealed-bid procurement auctions for selling a bundle of products through two lots with different sizes. We find that though the two auction mechanisms yield the same expected procurement costs to the buyer, other aspects of the two models, including the equilibrium bidding strategies as well as winning bidders’ ex post profits differ significantly. The

Bayesian-Nash equilibrium bid function also serves as a good predictor for bidding strategies in the lab.

2 - Revenue-monotonicity in Combinatorial Auctions

Marion Ott, RWTH Aachen University, Templergraben 64, Aachen,

52062, Germany, [email protected], Marissa Beck

Under what conditions does revenue in sealed-bid combinatorial auctions decrease when bids increase? We investigate revenue-(non)monotonicity in the

Vickrey Package Auction and Core-Selecting Auctions. In the Vickrey auction and in Minimum-Revenue Core-Selecting Auctions this counterintuitive phenomenon may appear even if we restrict bidders’ values to be submodular and under further restrictions.

3 - Hierarchical Model for Estimating Values in Sponsored Search

Eric Sodomka, PhD Candidate, Brown University, 115 Waterman

Street, 4th Floor, Providence, RI, United States of America, [email protected], Dustin Hillard, Sebastien Lahaie

We propose an approach to predicting sponsored search advertisers’ keyword values based on bidding behavior across the terms and campaigns in an account.

We propose an economically meaningful loss function which allows us to implicitly fit a hierarchical linear model for values given observables such as bids and click-through rates. Its predictive quality is evaluated on several highrevenue and high-exposure advertising accounts on a major search engine.

4 - An Empirical Analysis of Bidding Behavior in Overlapping

Online Auctions

Lin Hao, Assistant Professor, University of Notre Dame,

Mendoza College of Business, Notre Dame, IN, 46556, United

States of America, [email protected], Arvind Tripathi, Yong Tan

This paper studied bidding behavior in overlapping online auctions. We investigated how bidders choose their bids when multiple identical auctions are running concurrently. We identified factors affecting bidders’ decisions on which auction to bid. We also revealed how single-unit demand bidder’s bidding behavior differs from multi-unit demand bidder’s. We develop an empirical model to estimate those issues using the overlapping online auction data.

TA63

63- Borein B- Hyatt

Decision Making in Operational Contexts

Sponsor: Behavioral Operations

Sponsored Session

Chair: Enno Siemsen, University of Minnesota, 321 19th Ave. S,

Minneapolis, MN, United States of America, [email protected]

1 - Wholesale Pricing under Fairness Concerns

Julie Niederhoff, Syracuse University, Syracuse, NY,

United States of America, [email protected], Panos Kouvelis

Modeling analysis shows that, under fairness-based utility functions, the uncoordinated system is maximized if the supplier is generously fair-minded while the retailer is spiteful. Through a series of controlled experiments we consider these analytical predictions to study how suppliers set wholesale prices and how retailers set inventory levels in a newsvendor setting.

2 - Decision Bias in the Sequential Search for the Best Alternative:

A Laboratory Investigation

David Hall, PhD Candidate, Clemson University, 113 Daniel Ave.,

Seneca, SC, 29678, United States of America, [email protected],

Gulru Ozkan

How to optimally and sequentially search for the best alternative has been examined extensively in the extant New Product Development literature. The optimal search policy is to develop products with the highest net present value, often requiring radical innovation. Yet to the contrary, anecdotal evidence suggests that incremental innovation may be preferred. With this in mind, we ask: Do agent’s decision biases cause them to develop incremental products at the expense of radical products?

3 - The Anatomy of Newsvendor Order Decisions: Results from a

Task Decomposition Experiment

Yun Shin Lee, Judge Business School, University of Cambridge,

Trumpington Street, Cambridge, CB2 1AG, United Kingdom, [email protected], Enno Siemsen

We study the anatomy of a newsvendor order decision by decomposing it into point forecast, distribution forecast and service level decision and take a systematic approach to predict actual order behavior using these components. We also test whether we can improve order decisions by forcing decision makers to recognize the decomposed components.

4 - Decision Making and Cognition in Multi-Echelon Supply Chains:

An Experimental Study

Brent Moritz, Assistant Professor, Pennsylvania State University,

469 Business Building, University Park, PA, 16802, United States of America, [email protected], Arunchalam Narayanan

Supply chain performance often depends on the decisions of channel members.

Using a variant of the Beer Game, we find that the cognitive profile of the decision maker has an impact on performance, both for the supply chain and for a particular echelon. The effect is more pronounced in more complex problem environments. The findings have implications for supply chain design, education and industry employers.

282

INFORMS Phoenix – 2012

TA66

66- Ellis West- Hyatt

Data Driven Analytics in Support of Equipment

Health Management

Sponsor: Data Mining

Sponsored Session

Chair: Artur Dubrawski, Director, Auton Lab, Carnegie Mellon

University, 5000 Forbes Avenue NSH 3121, Pittsburgh, PA,

United States of America, [email protected]

1 - Data Overload: V-22 Osprey’s Foray into Big Data

Joseph Schmidley, V-22 ALE IPT Lead, Department of the Navy,

47123 Buse Road, B2272 Rm 146, Patuxent River, MD, 20670,

United States of America, [email protected]

After 100,000 flight hours, the V-22 program has started to leverage mass amounts of data from this sensor-rich tiltrotor platform. The requirement to collect/process/analyze aircraft data, with an increase in operational tempo, while continuing to grow in fleet size, the V-22 program must take action to capitalize on all of its data. We will review the V-22’s current challenges and the collaborative efforts through its transition from information technology to information capital.

2 - Detection of Anomalous Maintenance Patterns in Support of

Management of an Aviation Fleet

Norman Sondheimer, University of Massachusetts, Amherst, MA,

United States of America, [email protected],

Artur Dubrawski

Multivariate time series analysis of maintenance records to determine unusual patterns of behavior has been found effective in providing early warning of systematic failures of components. This presentation shows how these techniques were integrated into the management of aircraft fleets. The results can be generalized to fleets of other types of equipment.

3 - P-MATCH and QUBIT: Methods to Extract Critical Information in

Free Text for System Health Management

Anne Kao, Technical Fellow, Boeing, P.O. Box 3707,

MC 7L-43, Seattle, WA, 98124, United States of America, [email protected], Stephen Poteet, David Augustine

In data sources for system health management, free text data often provides critical information. Analysts need to extract detailed information with high recall from very noisy text data. We will use airplane health management to illustrate how to extract critical information from free text data and better support decision making in system health management. Our methods use text analytics and visual interactive techniques to leverage analysts’ knowledge to improve system health management.

4 - Intelligent Maintenance Decision Support Tool

Oscar Kipersztok, Technical Fellow, Boeing Research & Technology,

P.O. Box 3707, MC 7L-44, Seattle, WA, 98124,

United States of America, [email protected]

A decision support tool is presented to facilitate quick turn around at airplane maintenance and dispatch facilities. It is used for fault isolation and diagnosis. Its inputs are failure indication and its outputs are lists of probable causes and recommended maintenance actions. It is based on knowledge from troubleshooting and fault isolation manuals, and reliability data. It is aimed at reducing diagnosis errors and unnecessary component replacements to reduce operating costs.

TA67

67- Ellis East- Hyatt

Nanomanufacturing and Nanoinformatics I

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Qiang Huang, Associate Professor, University of Southern

California, 3715 McClintock Avenue, GER240, Los Angeles, CA, 90089,

United States of America, [email protected]

Co-Chair: Lijuan Xu, University of Southern California,

3715 McClintock Ave., GER 236, Los Angeles, CA, 90089,

United States of America, [email protected]

1 - Cross-Domain Model Building and Validation to Reinforce

Understanding of Nanowire Weight Kinetics

Li Wang, University of Southern California, 3715 McClintock Ave.,

RM 236, Los Angeles, CA, 90089, United States of America,

[email protected], Qiang Huang

Understanding nanostructure growth faces issues of limited data, lack of physical knowledge, and large process uncertainties. These issues result in modeling

TA68

difficulty because a large pool of candidate models almost fit the data equally well. In this talk, we present a new cross-domain model building and validation approach to address these challenges for scalable nanomanufacturing.

2 - Causal Inference for Scale-up in Nano-manufacturing

Tirthankar Dasgupta, Assistant Professor, Harvard University,

Science Center, 1 Oxford Street, Cambridge, MA, 02138-2901,

United States of America, [email protected],

Qiang Huang

Statistical inference for nanostructure growth models is typically drawn from small scale laboratory-level experiments. To scale up nanostructure growth, it is necessary to expand the scope of inference to super-populations which mimic actual manufacturing conditions. A framework for causal inference, based on potential outcomes, can provide a solution to this challenging problem.

3 - A Two-stage Statistical Approach to Modeling the Potential

Surface in Nanoquantification

Peter Qian, University of Wisconsin, Madison, WI,

United States of America, pe[email protected]

Quantifying the potential from nanomaterial under AFM scanning is challenged by the massive and complicated topography. A physical model is difficult to fully capture the relations between potential and topography. In this work, we propose a two-stage statistical approach to build the link between potential and topography and then explore the spatial dependence of the potential surface.

Better prediction performance of the new approach can be demonstrated by numerical study.

4 - Graphene Growth Process Modeling

Jian Wu, USC ISE, 3715 McClintock Avenue, GER 236,

Los Angeles, CA 90089-0193, United States of America, [email protected], Qiang Huang

Graphene growth on Cu foils shows four-lobed-symmetric shape and angle dependent growth velocity. We decompose the growth process and develop a modeling approach to describe the area coverage and geometric shape of graphene islands.

TA68

68- Suite 312- Hyatt

Simulation Techniques in Financial Risk Management

Sponsor: Financial Services Section

Sponsored Session

Chair: Wei Chen, Principal Product Manager, SAS Institute Inc.,

SAS Campus Dr., Cary, NC, 27512, United States of America,

[email protected]

1 - A Hybrid Simulation Methodology to Simulate Financial Risks

Jack Yang, Senior Research Statistician, SAS Institute, Inc, 100

SAS Campus Dr, Cary, NC, United States of America,

[email protected]

This paper introduces a simulation method that allows a combination of different simulation techniques in one simulation.It is especially suitable for a large financial system where each sub-group of risk factors can be simulated by using different data source and special modeling methods. The hybrid simulation technique accommodates all sub-models into a system and use copula to support the correlation structure.

2 - Efficient Monte Carlo Simulation in Counterparty Credit

Risk Measurement

Samim Ghamami, University of Southern California, Los Angeles,

CA, United States of America, [email protected], Bo Zhang

Counterparty credit risk measurement requires extensive use of Monte Carlo simulation. We analyze and comment on recent Monte Carlo counterparty credit risk methodologies being employed by financial institutions. We, also, develop an efficient Monte Carlo framework for estimation of average-type counterparty credit risk measures.

3 - A Dynamic Simulation Approach to Financial Risk Analysis

Sunny Zhang, Senior Risk Consultant, SAS, 100 SAS Campus Dr.,

Cary, NC, 27513, United States of America, [email protected],

Wei Chen, Jack Yang

We present a dynamic simulation approach with forward-looking assessment of market states. This new approach allows to control and select different simulation regimes and forecasting horizons as well as models based on the attributes of risk factors and portfolios. This approach can better forecast market states and capture extreme events so to improve various risk measures for VAR, stress testing and liquidity risk management, and assessment of economic capital across the entire enterprise.

283

TA69

INFORMS Phoenix – 2012

TA69

69- Suite 314- Hyatt

Optimization in Finance

Cluster: Optimization in Finance

Invited Session

Chair: Luis Zuluaga, Assistant Professor, Lehigh Univeristy,

200 West Packer Ave., Bethlehem, United States of America, [email protected]

1 - Stock Return Serial Dependence and Out-of-Sample

Portfolio Performance

Victor DeMiguel, Professor, London Business School, Regent’s

Park, London, NW1 4SA, United Kingdom, [email protected],

Raman Uppal, Francisco J. Nogales

We study whether investors can exploit stock return serial dependence to improve portfolio performance. First, we show that a vector autoregressive (VAR) model captures daily stock return serial dependence in a statistically significant manner. Second, we show that both arbitrage (zero-cost) portfolios and investment (positive-cost) portfolios that exploit serial dependence in stock returns outperform the traditional portfolios, even in the presence of transaction costs of up to 10 basis points.

2 - Tractable Asset-Liability Management under Time-Varying

Investment Opportunities

Dessislava Pachamanova, Associate Professor, Babson College,

Babson Park, MA, United States of America, [email protected]

We present models for asset liability management that are applicable to the management of pension funds and to the pricing of some kinds of structured financial products. The emphasis of the proposed approaches is on computational tractability and practical appeal. Numerical studies with real market data indicate that the proposed strategies have better worst-case performance than traditional stochastic approaches, as well as other desirable characteristics.

3 - Bounds on the Expected Payments of Insurance Instruments:

A Novel Computational Approach

Robert Howley, Lehigh University, 200 West Packer Avenue,

Harold S Mohler Lab, Bethlehem, PA, 18015-1582, [email protected], Luis Zuluaga, Robert Storer

It has been shown that single random variable semiparametric bounds can be solved using semidefinite programming (SDP). While the SDP approach was the foundation of many important advances in the literature, it has practical limitations. Here, we develop a method that uses standard optimization techniques, such as linear and first-order nonlinear programming. We also show that our approach allows the inclusion of extra information about the random variable (e.g., unimodality) in a simple way.

TA70

70- Suite 316- Hyatt

Interactive Advertising

Sponsor: Information Systems

Sponsored Session

Chair: Vibhanshu Abhishek, University of Pennsylvania, Phildelphia,

PA, United States of America, [email protected]

1 - Online Advertising Response Models with Multiple Creatives and Impression Histories

Michael Braun, Associate Professor of Marketing, Massachusetts

Institute of Technology, Sloan School of Management, 100 Main

St., E62-535, Cambridge, MA, 02139, United States of America, [email protected], Wendy Moe

Online advertising campaigns can consist of multiple ads with different creative content. We propose a model that evaluates the impact of each ad impression on both visitation and conversion behavior. Our model allows for the accumulation, decay, wearout and restoration of advertising effects. We show that online advertisers can increase the number of website visits and conversions by varying the creative content shown to an individual, according to that individual’s impression history.

2 - Information Asymmetry and Hybrid Advertising

De Liu, Associate Professor, Gatton College of Business and

Economics, University of Kentucky, Lexington, KY,

United States of America, [email protected], Siva Viswanathan

Pay-for-performance (P4P) advertising schemes such as pay-per-click have grown in popularity. Meanwhile the traditional pay-per-impression (PPI) scheme persists, and advertising providers start to offer a hybrid mix of PPI and P4P schemes. We examine optimal pricing schemes and demonstrate a trade-off between using P4P schemes to reveal superior quality and using PPI to minimize allocative inefficiencies. We shed light on several provider strategies, including increasing use of hybrid schemes.

3 - Bid Jamming

Shijie Lu, University of Southern California, 3660 Trousdale

Parkway, ACC 306, Marshall School of Business, Los Angeles, CA,

90089, United States of America, [email protected], Yi Zhu

Bid jamming is a practice in Generalized Second-Price (GSP) auctions that allows an advertiser at a lower position to bid one cent less than her competitor to accelerate the depletion of the competitor’s budget. This paper examines the impact of bid jamming on publisher’s revenue. The overall impact is found to be indefinite and is positively correlated with the competition intensity of the GSP auction. Several important managerial implications for both publishers and advertisers are discussed.

TA71

71- Suite 318- Hyatt

Improving Corporate Information Security through

Policy, Monitoring and Tools

Sponsor: eBusiness

Sponsored Session

Chair: Sam Ransbotham, Boston College, 140 Commonwealth Ave,

Chestnut Hill, MA, 02467, United States of America, [email protected]

1 - Is Corporate Social Responsibility Good for IS Security?

John D’Arcy, University of Delaware, Newark, DE,

United States of America, [email protected]

Using a combined dataset of security breaches (2005-2010) and ratings on over

30 different dimensions of corporate social responsibility (CSR), we explore the relationship between CSR and IS security breaches. Drawing on signaling theory, we hypothesize that internal breaches are less likely as CSR practices increase.

However, we also predict a positive relationship between CSR and external breaches. Contingency factors that influence these relationships will be discussed.

2 - How Does Continuous Auditing and Monitoring Impact

Employees’ Intentions to Commit Fraud?

France Bèlanger, Professor, Virginia Tech, 3007 Pamplin Hall,

Blacksburg, VA, 24061, United States of America, [email protected], Kathy Enget

Employee fraud has a potentially broad impact. Continuous auditing and monitoring (CA/CM) can be a valuable tool in this situation. This research explores CA/CM impacts on the fraud triangle via interviews and then with a vignette-based survey. The proposed vignettes focus on a manager needing to meet sales targets. With CA/CM, we expect to find lower intentions to commit fraud, lower perceived opportunities, no change in perceived pressures, and no change or lower ability to justify fraud.

3 - Impact of Anti-phishing Tool Performance on Attack

Success Rates

Ahmed Abbasi, University of Virginia, Rouss and Robertson Halls,

Charlottesville, VA, United States of America, [email protected], Mariam Zahedi, Yan Chen

Phishing website-based attacks continue to present significant problems for individual and enterprise-level security. While the performance of anti-phishing tools has improved, it remains unclear how effective such tools really are at protecting users. In this study, an experiment involving over 900 participants was used to evaluate the impact of anti-phishing tool performance on users’ ability to avoid different types of phishing threats.

284

INFORMS Phoenix – 2012

TA72

72- Suite 322- Hyatt

Computational Stochastic Optimization in Energy I

Sponsor: Computational Stochastic Optimization

Sponsored Session

Chair: Cosmin Petra, Argonne National Laboratory,

9700 S. Cass Avenue, Argonne, Il, 60439, United States of America, [email protected]

1 - Approximate Dynamic Programming for the Stochastic Control of a Wind Energy Storage System

Daniel Salas, Princeton University, 228 Sherrerd Hall, Princeton,

NJ, 08540, United States of America, [email protected],

Warren Powell

One of the main challenges in developing wind energy storage strategies is to maximize the allocation efficiency of the storage devices, which may be charged or discharged at each point in time. These devices are generally treated as simple black-box models which do not explicitly reflect the underlying physicochemical principles. ADP is proposed as a way to obtain a practical control policy for a more realistic battery system, by taking into account the uncertainty in supply and demand.

2 - Optimal Scheduling of Critical Peak Pricing Events with

Day-ahead Wind Energy Commitments

Xiaoxuan Zhang, IBM TJ Watson Research,

1101 Kitchawan Road, Route 134, Yorktown Heights, NY, 10598,

United States of America, [email protected], Bo Zhang

We investigate the optimal scheduling of critical peak pricing from the perspective of a load serving entity who has wind energy to commit. The goal is to minimize the total cost for the whole planning horizon, including energy purchasing cost, revenue from the critical peak pricing, wind sales, and imbalance penalties due to wind over- and under-commitments. We propose a multi-stage stochastic mixed integer nonlinear programming model and present an optimal stopping view of this model.

3 - Emulating Stochastic Programming Effects in the Unit

Commitment Problem

Boris Defourny, Princeton University, ORFE, Sherrerd Hall,

Charlton St., Princeton, NJ, 08544, United States of America, [email protected], Hugo Simao, Warren Powell

We study modifications of the deterministic unit commitment problem for emulating the behavior of a multistage stochastic model without optimizing over multiple scenarios. To do that we distort the forecasted demand and wind energy seen by the unit commitment model, and we introduce optimization variables and constraints for representing special kinds of reserve requirements, tuned by simulation or by solving auxiliary stochastic programs.

4 - Stochastic Optimization of Power Grid on

High-performance Computers

Cosmin Petra, Argonne National Laboratory, 9700 S. Cass Avenue,

Argonne, Il, 60439, United States of America, [email protected],

Mihai Anitescu, Miles Lubin

We present a scalable framework for solving stochastic programming problems, with application to the optimization of power grid energy systems with uncertain supply and demand. Our framework, PIPS, has parallel capabilities for both continuous and discrete stochastic optimizations problems. We will also discuss the computational results obtained on “Intrepid” Blue Gene/P system at Argonne when solving unit commitment problems with billions of variables.

TA73

73- Suite 324- Hyatt

Topics on American Options

Cluster: Quantitative Finance

Invited Session

Chair: Daniel Mitchell, University of Texas, McCombs School of

Business, 1 University Station, B6000, Austin, TX, 78712,

United States of America, [email protected]

1 - Boundary Evolution Equations for American Options

Daniel Mitchell, University of Texas, McCombs School of Business,

1 University Station, B6000, Austin, TX, 78712, United States of

America, [email protected],

Kumar Muthuraman, Jonathan Goodman

We consider the problem of finding optimal exercise policies for American options in constant and stochastic volatility. We derive and use boundary evolution equations that characterize the evolution of the optimal exercise boundary. Using these boundary evolution equations we show how one can construct very efficient computational methods for pricing American options.

2 - American Option Pricing for Subordinate Diffusions

Lingfei Li, Assistant Professor, The Chinese University of Hong

Kong, Shatin, New Territories, Hong Kong-PRC, [email protected], Vadim Linetsky

We develop an eigenfunction expansion approach for pricing American options under models based on subordinate diffusions (i.e. jump-diffusions or pure jump processes obtained by time changing diffusions with Levy subordinators). As an application we consider American option pricing under the SubOU-based commodity model with mean-reverting jumps.

3 - Pricing and Static Hedging of American Options under the

Jump to Default Extended CEV Model

Jose Carlos Dias, ISCTE-IUL Business School, Complexo

INDEG/ISCTE, Av. Prof. Anibal Bettencourt, Lisboa, Portugal, [email protected], Joao Pedro Ruas, Joao Pedro Nunes

This paper prices American options through the static hedge approach (SHP) and extends the literature in two directions. First, American options are priced under the CEV model of Cox (1975) through the SHP procedure offered by Chung and

Shih (2009), but now extended for any elasticity parameter (beta) of the CEV process. Second, the SHP approach is adapted to the JDCEV model of Carr and

Linetsky (2006), and American options on defaultable equity are also priced.

Tuesday, 11:00am - 12:30pm

TB01

TB01

01- West 101- CC

Nonlinear Integer Optimization and Applications

Sponsor: Optimization/Global Optimization

Sponsored Session

Chair: Serdar Karademir, University of Pittsburgh, 1048 Benedum Hall,

Pittsburgh, United States of America, [email protected]

1 - Irregular Polyomino Tilings via Integer Programming

Serdar Karademir, University of Pittsburgh, 1048 Benedum Hall,

Pittsburgh, United States of America, [email protected],

Oleg Prokopyev

Polyominoes are created by connecting unit squares along an edge. Recently, irregular polyomino tilings have been used in the design of array antennas to improve their imaging capabilities. We formulate the problem as a nonlinear exact set covering model where irregularity is measured using the information theoretic entropy concept. A column generation approach along with novel branching and lower bounding schemes is proposed. We also provide efficient heuristics and approximation algorithms.

2 - On a Class of Nonlinear Bilevel Knapsack Problems

Behdad Beheshti, University of Pittsburgh, 1048 Benedum Hall,

Pittsburgh, PA, 15261, United States of America, bek[email protected],

Oleg Prokopyev, Osman Ozaltin

We study a bilevel knapsack problem where the leader’s objective is in a nonlinear form. We discuss related theoretical computational complexity issues.

We also present an exact enumeration approach based on a specialized branchand-backtrack algorithm. The performance of the proposed approach is evaluated on a large number of instances.

3 - Improving the Quality of the Assignment of Students to

First-Year Seminars

Richard Forrester, Associate Professor of Mathematics, Dickinson

College, College and Louther Streets, Carlisle, PA, 17013,

United States of America, [email protected], Kevin Hutson,

Thanh To

Many institutions of higher education have a First-Year Seminar Program. At

Dickinson College, students are required to select six seminars from a list of 42 possibilities. Using standard optimization software, we assign students to seminars so as to balance both the gender and number of international students in the seminars. In addition, we use Monte Carlo simulation to study how the number of seminars each student is required to select affects the likelihood that a feasible assignment exists.

4 - Nonlinear Mixed Integer Programming in Power

Network Optimization

Pavlo Krokhmal, University of Iowa, 3131 Seamans Center,

Iowa City, IA, 52242, United States of America, [email protected], Bo Sun

We consider the problem of optimizing power load distribution in networks, which reduces to non-linear mixed integer second order cone programming model. An algorithm based on cutting plane method for solving polyhedral approximations of the relaxed problem is proposed. Numerical case studies are presented.

285

TB02

INFORMS Phoenix – 2012

TB02

02- West 102 A- CC

Advice Taking and Opinion Aggregation

Sponsor: Decision Analysis

Sponsored Session

Chair: Jack Soll, Associate Professor, Duke University, 100 Fuqua Drive,

Durham, NC, 27708, United States of America, [email protected]

1 - Combining the Intuitive and Analytic Mind

Jack Soll, Associate Professor, Duke University, 100 Fuqua Drive,

Durham, NC, 27708, United States of America, [email protected],

Rick Larrick

We show the benefits of having the same individual make a rapid intuitive judgment and a slower deliberate judgment, and then averaging them. There is an asymmetry, however, such that the intuitive-analytic sequence allows for greater “refreshing” and bracketing than the reverse sequence. Similar benefits occur when combining strategies across people. We discuss the implications for debiasing strategies.

2 - Double-sided Utility Copulas

Ali Abbas, Associate Professor, University of Illinois at Urbana-

Champaign, 104 S. Mathews Ave., Urbana, IL, 61822,

United States of America, [email protected]

A double-sided utility copula matches the conditional utility assessments at all boundary values of the domain of the attributes and allows for an additional degree of freedom to vary the trade-off assessments among them. We should how to construct such utility copula functions and discuss methods for their assessment.

3 - The Wisdom of Small Crowds of Probability Forecasters

Yael Grushka-Cockayne, Assistant Professor, University of Virginia,

Darden School of Business, 100 Darden Blvd, Charlottesville, VA,

22903, United States of America, [email protected],

Albert Mannes, Casey Lichtendahl

Recent evidence suggests that the accuracy of the average point forecast increases rapidly in crowd size initially and then slowly thereafter. We study this effect in the context of probability forecasting. We find that the accuracy of the average probability forecast increases up to a certain crowd size and then decreases thereafter.

TB03

03- West 102 B- CC

Games, Decisions, and Descriptive Models

Beyond Rationality

Sponsor: Decision Analysis

Sponsored Session

Chair: Seth Guikema, Assistant Professor, Johns Hopkins University,

313 Ames Hall, Department of Geog & Env. Engineering, Baltimore,

MD, 21218, United States of America, [email protected]

1 - Beyond Rationality: Some Examples of Descriptive Models of

Behavior in Decision Settings

Seth Guikema, Assistant Professor, Johns Hopkins University, 313

Ames Hall, Department of Geog & Env. Engineering, Baltimore,

MD, 21218, United States of America, [email protected]

Decisions must be made in the presence of other intelligent actors in many settings such as terrorist risk management and environmental regulation and enforcement. Model are often used to describe the actions of these other players, and there is an increasing interest in behavioral models of these other players.

This talk gives an overview of recent work in this area and sets the stage for the three research talks in the session.

2 - The Matching Law: Application of a Psychological Model to a

Management Problem

Philip Leclerc, Virginia Commonwealth University, P.O. Box

843083, Richmond, VA, 23284, United States of America, [email protected], Laura McLay, Jason Merrick

Many optimization models require that human behavior be modeled. We introduce Herrnstein’s Matching Law, a model of choice behavior developed in operant psychology. We explore a class of non-convex optimization problems created by introducing the Matching Law into a preexisting optimization problem, and demonstrate how to approximately solve a particular class of such problems using separable programming. We apply our method to an assignment problem.

3 - Rationality Bias in Water Resource Management

Stefanie M. Falconi, PhD Student, Johns Hopkins University,

3400 N. Charles St., Ames 313, Baltimore, MD, 21218,

United States of America, [email protected]

In the water resource management field the rationality axioms have been central to systems analysis models. Empirical studies of water management in real political context point to the oversimplifications of these assumptions. What is the implication of this bias and why does it pose challenges to how water gets managed? This talk will explore several case studies to highlight the inconsistencies between the rationality axioms and the objectives of integrative water resource management.

4 - Agent-based Simulation of Learning Adversaries in

Defender-attacker Games

Emily Zechman, Assistant Professor, North Carolina State

University, CB 7908, Raleigh, NC, 27695, United States of

America, [email protected], Seth Guikema

Advanced analytical methods typically model defender-attacker games by assuming a static decision-making process for players. Over a series of plays and pay-offs, however, both defenders and attackers may adapt their decision-making strategy and their assumptions about the adversary’s decision-making strategy.

This research couples agent-based modeling with game theory to simulate the evolution of a game and pay-offs as players learn.

TB04

04- West 102 C- CC

Data Envelopment Analysis II

Cluster: Data Envelopment Analysis

Invited Session

Chair: Joe Zhu, Professor, Worcester Polytechnic Institute, 100 Institute

Rd, Worcester, MA, 01609, United States of America, [email protected]

1 - Improving DEA Efficiency under Multiple Perspectives by

Game Theory

Xiaopeng Yang, University of Toronto, Centre for Management of

Technology, Toronto, Canada, [email protected],

Joseph Paradi, Hiroshi Morita

In the current research, each perspective stands for a classification of input/output for the attributes of DMU. Different perspectives have distinctive preferences even for the same attribute of DMU. In order to improve the DEA efficiency of a DMU under such a conflictive circumstance, we propose a new calculation method using game theory. A numerical case study of banking systems is also given to show the calculation results of our approach.

2 - Use of DEA Cross-efficiency Evaluation in Portfolio Selection

Sungmook Lim, Associate Professor, Korea University,

Jochiwon-eup, Yeongi-gun, Chungnam, 339700, Korea, Republic of, [email protected], Joe Zhu

We propose a novel use of DEA cross-efficiency evaluation in portfolio selection.

In addition to average cross-efficiency scores, we suggest to examine the variations of cross-efficiencies, and to incorporate the two statistics of crossefficiencies into the mean-variance formulation of portfolio selection. Two benefits can be attained by this approach; selection of robust portfolios and alleviation of the so-called “ganging together” phenomenon of DEA crossefficiency evaluation.

3 - Network DEA Pitfalls

Joe Zhu, Professor, Worcester Polytechnic Institute, 100 Institute

Rd, Worcester, MA, 01609, United States of America, [email protected], Yao Chen, Wade Cook, Chiang Kao

Network DEA models been developed to examine the efficiency DMUs with internal structures. Pitfalls in network DEA are discussed with respect to the determination of divisional efficiency, frontier type, and projections.

286

INFORMS Phoenix – 2012

TB05

05- West 103 A- CC

Decision Analysis Approaches and Predictive

Modeling to Managing Uncertainty in Manufacturing and Service Systems Design & Operations

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Zhenyu (James) Kong, Assistant Professor, Oklahoma State

University, 322 Engineering North, Stillwater, OK, 74078,

United States of America, [email protected]

Co-Chair: Martin Wortman, Professor, Texas A&M University, [email protected]

1 - Modeling and Prediction of Degradation Profiles in Tissueengineered Scaffold Fabrication

Li Zeng, Assistant Professor, The University of Texas at Arlington,

500 West First Street, Arlington, TX, 76019,

United States of America, [email protected], Xinwei Deng

Tissue-engineered scaffolds is an important bio-manufacturing products, which plays a critical role in the development of engineered tissues/organs. One important quality characteristic of tissue-engineered scaffolds is their degradation profile in human bodies. This study builds statistical models for the degradation profiles and considers predicting degradation of scaffolds under given experimental settings based on the model.

2 - Chemical Mechanical Planarization (CMP) Process Monitoring using Evolutionary Clustering Analysis

Omer Beyca, Research Assistant, Oklahoma State University,

Stillwater, OK, 74074, United States of America, [email protected], Zhenyu (James) Kong,

Satish Bukkapatnam

Process monitoring for Chemical Mechanical Planarization (CMP) is extremely crucial to detect process defects and ensure high quality of polished wafers in semiconductor industry. It is challenging due to the non-linear and nonstationary nature. We proposed a novel Recurrent Nested Dirichlet Process

(RNDP) model to tackle this challenge by dynamically constructing mixture of

Gaussian distributions in a Markovian manner. Our case studies demonstrated the effectiveness of the proposed method.

TB06

06- West 103 B- CC

Monte Carlo Methods in Finance

Sponsor: Simulation

Sponsored Session

Chair: Liming Feng, Assistant Professor, University of Illinois at Urbana-

Champaign, 104 S Mathews Ave, Urbana, IL, 61801,

United States of America, [email protected]

1 - American Option Sensitivities Estimation via a Generalized

IPA Approach

Nan Chen, Assistant Professor, The Chinese University of Hong

Kong, Shatin, N.T., Hong Kong, Hong Kong-PRC, [email protected], Yanchu Liu

We develop efficient Monte Carlo methods for estimating American option sensitivities. A generalized infinitesimal perturbation analysis (IPA) approach introduced to resolve the difficulty caused by discontinuity of the optimal decision with respect to the underlying parameter. In contrast to the conventional IPA method, we find that the pathwise differentiability is not necessary for the application of our generalized IPA. Numerical implementation and error analysis are also discussed.

2 - Denoising Monte Carlo Sensitivity Estimates

Kyoung-Kuk Kim, Assistant Professor, KAIST, 291 Daehak-ro,

Yuseong-gu, Daejeon, 305-701, Korea, Republic of, [email protected], Wanmo Kang, Hayong Shin

Monte Carlo methods are popular in academia and industry. We are interested in improving sensitivity estimates obtained from MC experiments with respect to given parameter values, motivated by, but not restricted to, financial applications.

Denoising and interpolation methods, which have been used for a long time in many different areas, are proposed in a new form which is quadratic, easy to implement, and tailored to our objectives. This heuristic approach is supported by numerical experiments.

TB07

3 - Simulation-based Maximum Likelihood Estimation for Levydriven OU Stochastic Volatility Models

JianQiang Hu, Professor, Fudan University, Department of

Management Science, Shanghai, 200433, China, [email protected], Yijie Peng, Michael Fu

We study the problem of estimating parameters for Levy-Driven Ornstein-

Uhlenbeck stochastic volatility models, which has a wide range of applications in finance, econometrics, and statistics. We consider the problem in the framework of maximum likelihood estimation and use various techniques in simulation optimization, including conditional Monte Carlo sensitivity estimation and sequential Monte Carlo method for hidden Markov processes. We present numerical results to validate out method.

4 - Simulating from Analytic Characteristic Functions

Liming Feng, Assistant Professor, University of Illinois at Urbana-

Champaign, 104 S Mathews Ave., Urbana, IL, 61801,

United States of America, [email protected], Zisheng Chen

Analytic characteristic functions naturally arise in financial engineering applications. We explore the analyticity of such characteristic functions and propose simple but accurate inversion schemes. They admit explicit error estimates that allow us to derive explicit bounds for the estimation bias when the inverse transform method is used to simulate from such analytic characteristic functions.

TB07

07- West 104 A- CC

Data Mining in Medical Decision Making and

Bioinformatics Applications

Sponsor: Data Mining

Sponsored Session

Chair: Kamran Paynabar, University of Michigan, Ann Arbor, MI,

United States of America, [email protected]

1 - Bayesian Analysis of Multiway Table of Genetic Effects:

A Model Comparison Approach

Xiaoquan Wen, University of Michigan, 1415 Washington Heights

Road, Ann Arbor, MI, 48109, [email protected]

We consider the inference problem on unobserved quantities that can be represented by multiway tables. We focus on Bayesian model comparisons in complex linear systems, in which we show multiway tables are naturally formed by regression coefficients. We specify prior distributions to enable efficient information sharing across and within different dimensions of multiway tables.

Analytic results of Bayes Factors and their approximations are derived and their various properties are discussed.

2 - Supervised Link Prediction in Biological Networks using

Penalized, Multi-Mode ERGMs

Ali Shojaie, University of Washington, F650 Health Sciences Bldg,

Department of Biostatisitcs, Seattle, WA, 98109,

United States of America, [email protected]

Available methods for inferring biological networks often focus on a single type of interaction among genes, which can results in incomplete information and high false positive rates. To address this limitation, we propose a multi-mode network model, based on exponential random graph models (ERGM), which represents different types of interactions, and provides a natural method of estimating networks based on known interactions and diverse sources of biological data.

3 - Self-organized Signal Quality Control in

Cardiorespiratory Monitoring

Hui Yang, Assistant Professor, University of South Florida, Tampa,

FL, 33620, United States of America, [email protected]

Telemedicine is critical to ensure the timely delivery of health care to patients, especially those who live in the rural areas. However, there are a number of uncertainty factors inherent to the telehealthcare, for e.g., the personnel with minimal training. This paper presents our efforts to integrate multiscale recurrence analysis with self-organizing map for controlling the signal quality.

The efficacy and robustness of this approach are validated using real-world telemedicine recordings.

4 - Decision Making about Surgical Treatment for Rotator Cuff Tear

Patients Based on Patients’ Clinical

Kamran Paynabar, University of Michigan, Ann Arbor, MI,

United States of America, [email protected], Bruce Miller,

Richard Hughes, Judy Jin

Treatment of patients with rotator cuff tears usually starts with a period of physical therapy. Patients who fail to respond to physical therapy are treated surgically. Identifying such patients, in initial clinical examination would help reduce the time and costs of treatment by skipping physical therapy. The goal of this study is to develop a regularized classification model for predicting the necessity of surgery based on the results of a patient’s clinical examination.

287

TB08

INFORMS Phoenix – 2012

TB08

08- West 104 B- CC

Joint Session PPSN/HAS: Optimization in

Health Policy

Sponsor: Public Programs, Service and Needs & Health

Applications Society

Sponsored Session

Chair: David Hutton, University of Michigan, 1415 Washington

Heights, SPH II: M3525, Ann Arbor, MI, 48109,

United States of America, [email protected]

1 - Pricing Strategies for Combination Pediatric Vaccines Based on the Lowest Overall Cost Formulary

Banafsheh Behzad, PhD Candidate, University of Illinois at

Urbana-Champaign, 117 Transportation Building,

104 S. Mathews Ave., Urbana, IL, 61801, United States of America, [email protected], Sheldon Jacobson, Edward Sewell

This paper analyzes pricing strategies for pediatric combination vaccines by comparing the lowest overall cost formularies for a fixed cost of injection. Three pharmaceutical companies compete over the sale of pediatric vaccines. The main contribution of the paper is to provide the lowest overall cost formularies for a fixed cost of an injection for different prices of Pentacel. The resulting analysis shows that Pentacel could have been more competitively priced compared to the

Pediarix.

2 - The Balance Optimization Subset Selection (BOSS) Model for

Causal Inference

Jason Sauppe, University of Illinois, Dept. of CS, Urbana, IL,

United States of America, [email protected], Sheldon Jacobson,

Edward Sewell, Alexander Nikolaev

Scientists in many fields rely on observational (non-random) data to estimate the effects of a treatment. To control for potential sources of bias in the estimate, a matching procedure is used. The success of the matching process is assessed using a balance measure. The Balance Optimization Subset Selection model eliminates the matching step and optimizes the balance measure directly. This has the potential to lead to more informed decisions about treatment effectiveness.

3 - Strategic Health Care Workforce Management

Alejandro Toriello, Assistant Professor, Industrial and Systems

Engineering, University of Southern California, 3715 McClintock

Ave., GER 240, Los Angeles, CA, 90089, United States of America, [email protected], Mariel Lavieri

We propose a long-term workforce management model for a large health care system with arbitrary deterministic non-decreasing demand, using infinite linear programming. The model monitors hiring, training and promotion of a type of health care worker across the entire system. We provide a series of conditions the system can satisfy to ensure that one-period lookahead policies are optimal. We also give a sensitivity analysis of the cost impact of various system parameters, such as demand.

4 - Using Antivirals for Prevention or Treatment During an Influenza Pandemic

Benjamin Armbruster, Assistant Professor, Northwestern

University, Evanston, IL, 60208, United States of America, [email protected], David Hutton

Many countries are stockpiling antivirals for prevention and treatment of pandemic influenza. We analyze an infectious disease model to determine the optimal allocation of the stockpile between prevention and treatment. We characterize the optimal solution analytically. Allocating current national antiviral stockpiles to treatment are likely to minimize population mortality. But, with much larger stockpiles, using antivirals for prevention may be optimal.

TB09

09- West 105 A- CC

A History of Multiple Criteria Decision Making

Sponsor: Multiple Criteria Decision Making

Sponsored Session

Chair: Stanley Zionts, Distinguished Professor Emeritus, School of

Management, State University of New York, Buffalo, NY, 14260,

United States of America, [email protected]

1 - A History of Multiple Criteria Decision Making

Stanley Zionts, Distinguished Professor Emeritus, School of

Management, State University of New York, Buffalo, NY, 14260,

United States of America, [email protected]

This presentation is based on a book, Multiple Criteria Decision Making: From

Early History to the 21st Century, published by World Scientific, Singapore, 2011 authored by Murat Köksalan, Jyrki Wallenius, and myself. It covers multiple criteria decision making and related developments from the earliest known works to work of the twentieth century.

TB10

10- West 105 B- CC

Joint Session OPT/ENRE: Stochastic Programming and Power System Optimization

Sponsor: Optimization/Stochastic Programming & Energy, Natural

Res & the Environment/Energy

Sponsored Session

Chair: Yongpei Guan, Associate Professor, University of Florida,

Department of Industrial and Systems, Engineering 303 Weil Hall,

Gainesville, FL, 32608, United States of America, [email protected]

1 - Multi-area Optimal Power Flow with Reconfigurable

Transmission Network

Cong Liu, Computational Engineer-Energy Systems, Argonne

National Laboratory, 9700 South Cass Ave., Building 221,

Argonne, IL, 60439, United States of America, [email protected],

Jianhui Wang

The model separates the whole network into sub-networks by graph partition.

Lagrangian multipliers and quadratic penalty terms will be introduced to relax the boundary power flow variables of tie-lines between two adjacent areas. Then the dual problem of the original problem will be decomposed into several subproblems associated with each area. The program can stop once the gap between the upper and lower bounds satisfies a small predefined value.

2 - Two-stage Robust Optimization for N-k Contingencyconstrained Unit Commitment

Qianfan Wang, University of Florida, 303 Weil Hall, Gainesville,

FL, United States of America, [email protected], Jean-Paul Watson,

Yongpei Guan

A two-stage robust optimization approach is presented to solve the general N-k contingency-constrained unit commitment (CCUC) problem. Both generator and transmission line contingencies are considered in this research. We propose a decomposition framework to enable tractable computation. In our algorithm, the master problem makes unit commitment decisions and the subproblem inspects the worst-case contingency scenarios. The computational results verify the effectiveness of the proposed approach.

3 - Cutting Planes for The Multi-stage Stochastic Unit

Commitment Problem

Ruiwei Jiang, University of Florida, 411 Weil Hall, University of

Florida, Gainesville, FL, 32611, United States of America, [email protected], Jean-Paul Watson, Yongpei Guan, Ming Zhao

Motivated by the intermittency of renewable energy, we propose a multi-stage stochastic integer programming model in this talk to address unit commitment problems under uncertainty, for which we construct several classes of strong inequalities by lifting procedures to strengthen the original formulation. Our preliminary computational experiments show encouraging results.

TB11

11- West 105 C- CC

Optimization Algorithms

Sponsor: Optimization/Nonlinear Programming

Sponsored Session

Chair: Daniel Robinson, Johns Hopkins University, 100 Whitehead Hall,

3400 N. Charles Street, Baltimore, MD, 21218, United States of

America, [email protected]

1 - An Adaptive Augmented Lagrangian Method

Daniel Robinson, Johns Hopkins University, 100 Whitehead Hall,

3400 N. Charles Street, Baltimore, MD, 21218, United States of

America, [email protected], Hao Jiang, Frank Curtis

We present a method for solving constrained optimization problems based on the augmented Lagrangian. In contrast to traditional augmented Lagrangian methods that slowly update multiplier estimates and an associated penalty parameter, our method updates both during each iteration. A strength of our method is the ability to approximately solve each subproblem and to efficiently update the penalty parameter to promote fast local convergence. Preliminary numerical results will be presented.

288

2 - On the Global Linear Convergence of Alternating

Direction Method

Wei Deng, Rice University, 6100 Main St., MS-134, Rice U,

CAAM Dept, Houston, TX, 77005, United States of America, [email protected], Wotao Yin, Yin Zhang

The alternating direction method has proven to be effective for solving many optimization problems, and has wide applications in areas such as compressive sensing, signal and image processing, machine learning, computer vision and statistics. We give sufficient conditions for global linear convergence of the alternating direction method where the iterates are allowed to be computed approximately using proximal methods.

3 - Sparse Approximation via Penalty Decomposition Methods

Zhaosong Lu, Professor, Simon Fraser University, 8888 University

Drive, Burnaby, BC, V5A 1S6, Canada, [email protected],

Yong Zhang

We propose penalty decomposition (PD) methods for general $l_0$ minimization.

Under suitable assumptions, we establish that any limit point of the sequence generated by the PD methods satisfies the first-order optimality conditions of the problems. Furthermore, for the problems in which the $l_0$ part is the only nonconvex part, we show that such a limit point is a local minimizer.

Computational results show that our methods generally outperform the existing methods.

TB12

12- West 106 A- CC

Joint Session Optimization IP/ICS: Constraint

Programming Methodology and Applications I

Sponsor: Optimization/Integer Programming & Computing Society

Sponsored Session

Chair: Willem-Jan van Hoeve, Carnegie Mellon University,

5000 Forbes Avenue, Pittsburgh, PA, United States of America, [email protected]

1 - Resolution of the Package Server Location Problem

Jean-Charles Regin, Professor, University Nice-Sophia Antipolis,

I3S, 2000, Route des Lucioles, Les Algorithmes, Euclide B, BP 121,

Sophia Antipolis, 06903, France, [email protected],

Mohamed Rezgui, Arnaud Malapert, Jean Parpaillon, Yvan Manon

We solve the package server location problem. A number of package servers are to be located at nodes of a network. Demand for services of these servers is located at each node, and a subset of nodes are to be chosen to locate one or more package servers in each. Each customer is associated to a package. The objective is to minimize the number of package servers while maximizing quality of services related to the efficiency and the reliability of the broadcast of packages to all customers.

2 - Retail Scheduling for Profit

Louis-Martin Rousseau, Polytechnique Montréal, CP 6079 Succ.

Centre-ville, Montrèal, Canada, [email protected],

Marc Joliveau, Nicolas Chapados, Pierre L’Ecuyer

We how that shift scheduling can be used to maximize operation profit rather than minimizing cost in retail stores. We evaluate the performance of several MIP models and Constraint Programming models, on both a daily and a weekly horizon.

3 - Joint Assessment and Restoration of Power Systems

Pascal Van Hentenryck, Professor, University of Melbourne,

Melbourne, Australia, [email protected], Carleton Coffrin,

Nabeel Gillani

This paper studies the joint damage assessment and recovery of the power infrastructure after a natural disaster has occurred. Earlier work in this area proposed an optimization algorithm for the recovery phase, assuming that the infrastructure damage was known precisely. This paper lifts this assumption and proposes online stochastic optimization approaches that use constraint programming to solve the deterministic problems.

4 - Constraint Restriction in Constraint Satisfaction Problems

Merav Aharoni, IBM Research-Haifa, Haifa University Campus,

Haifa, 31905, Israel, [email protected], Wesam Ibraheem,

Yehuda Naveh, Elena Tsanko

We study the usage of constraint restriction in the process of systematic solving of

Constraint Satisfaction Problems. Constraint restriction combines propagation with an arbitrary narrowing of the domains. It can narrow several domains simultaneously, while preserving consistency. We show how to incorporate constraint restriction into a MAC based backtrack search algorithm without compromising the algorithm completeness. We demonstrate several cases where it is useful to apply this technique.

INFORMS Phoenix – 2012

TB14

TB13

13- West 106 B- CC

Solving Hard Problems in Conic Optimization I

Sponsor: Optimization/Linear Programming and Complementarity

Sponsored Session

Chair: Akiko Yoshise, University of Tsukuba, 1-1-1 Tennnoudai,

Tsukuba, 305-8573, Japan, [email protected]

1 - Computation of Facial Reduction

Hayato Waki, Associate Professor, Kyusyu University,

744 Motooka, Nishi-ku, Fukuoka, 819-0395, Japan,

[email protected], Masakazu Muramatsu

Facial reduction algorithm (FRA) generates a smaller conic optimization problem by exploiting the degeneracy in a given conic optimization problem. However, the computation is comparable to solving the original problem. In addition, the resulting problem may lose the sparsity. In this talk, we show that one can apply

FRA effectively by using structure in the original problem. For this, we present some examples. This is a jointwork with Professor Masakazu Muramatsu (UEC).

2 - Determining Protein Structures by Semidefinite Programming

Nathan Krislock, INRIA Grenoble Rhòne-Alpes, 655 Avenue de l’Europe, Montbonnot, Saint-Ismier Cedex, 38334, France, [email protected], Babak Alipanahi, Ali Ghodsi,

Henry Wolkowicz, Logan Donaldson, Ming Li

Euclidean distance matrix methods based on semidefinite programming (SDP) are a natural approach for protein structure determination. Our contribution is a new

SDP formulation that overcomes the difficulty of the high complexity of SDP solvers by using facial reduction to reduce the problem to approximately one quarter the size of the original problem. The reduced SDP can be solved approximately 100 times faster, and it is more resistant to numerical problems from erroneous and noisy distances.

3 - Numerical Computation of a Facial Reduction Algorithm for

Doubly Nonnegative Optimization Problems

Mirai Tanaka, PhD. Student, Tokyo Institute of Technology,

2-12-1-W9-60 Ookayama, Meguro-ku, Tokyo, 152-8552, Japan, [email protected], Kazuhide Nakata, Hayato Waki

A facial reduction algorithm works for conic optimization problems and generates an equivalent conic optimization problem that has interior feasible solutions in finitely many iterations. This algorithm is great in theory but it has been considered to be difficult use in practice. In this talk, we introduce a numerical computation method of the facial reduction algorithm for doubly nonnegative optimization problems.

TB14

14- West 106 C- CC

Practical Batching and Supply Chain

Scheduling Problems

Cluster: Scheduling and Project Management

Invited Session

Chair: Zhi-Long Chen, Professor, University of Maryland,

Robert H. Smith School of Business, College Park, MD, 20742,

United States of America, [email protected]

1 - Integrated Charge Batching and Casting Width

Selection at Baosteel

Zhi-Long Chen, Professor, University of Maryland, Robert H.

Smith School of Business, College Park, MD, 20742,

United States of America, [email protected],

Gongshu Wang, Lixin Tang

We study an integrated charge batching and casting width selection problem arising in the continuous casting operation of the steelmaking process. For the general problem, we develop a column generation based branch and bound

(B&B) solution approach to obtain optimal solutions. We also consider a frequently occurring case of the problem where each steel grade is incompatible with any other grade. For this special case, we develop a two-level polynomialtime algorithm to obtain optimal solutions.

289

TB15

2 - A Branch-and-Price-and-Cut Algorithm for Batch Annealing in

Steel Industry

Ying Meng, Northeastern University, The Logistics Institute,

Shenyang, 110004, China, [email protected], Lixin Tang,

Zhi-Long Chen

We consider a batching decisions problem arising in the batch annealing operations and propose a branch-and-price-and-cut solution algorithm. To strengthen the upper bound, in addition to new columns, some valid inequalities are added to the formulation in every iteration of the column generation process.

A variable reduction strategy is also proposed to accelerate the algorithm. For some special cases, polynomial-time solution methods are proposed to solve the problem optimally.

3 - Production Scheduling with Two-modal Transportation

Consideration

Feng Li, Northeastern University, The Logistics Institute,

Shenyang, 110004, China, [email protected], Zhi-Long Chen,

Lixin Tang

Motivated by application in the steel industry, we study an integrated scheduling model of production and two-modal transportation. A set of jobs are first processed in a processing facility, then delivered to the port by trucks, and finally delivered to the customers by ships. We study various classes of problems. For each of the problems studied, we provide an efficient exact algorithm or a proof intractability accompanied by a heuristic algorithm with worst case analysis.

4 - An Improved Algorithm for a Single Machine Scheduling

Problem with Delivery Coordination

Weiya Zhong, Professor, Shanghai University, Department of

Mathematics, Shanghai, 200444, China, [email protected]

We consider a scheduling problem with delivery coordination where a manufacturer needs to determine a joint schedule for order processing and order delivery. Order delivery is carried out by a third party logistics provider. The objective is to minimize total delivery cost such that all the orders are delivered before their due dates. We present an algorithm with worst case ratio of 5/3, which improves the known ratio of 2 of an existing algorithm in the literature.

INFORMS Phoenix – 2012

TB16

16- West 207- CC

Modeling and Analysis of Perishable Inventory

Systems with Random Input

Cluster: Tutorials

Invited Session

Chair: David Perry, University of Haifa, Department of Statistics, Haifa,

26250, Israel, [email protected]

1 - Perishable Inventory Systems with Random Replenishments

David Perry, University of Haifa, Department of Statistics, Haifa,

26250, Israel, [email protected], Wolfgang Stadje

A guide to perishable inventory systems (PIS’s) that are refilled by randomly arriving items and not by ordering decisions is introduced. The literature on this class of PIS’s (for which a blood bank or an organ transplantation center are prominent examples) is sparse. The survey starts with the pioneering work on a prototype model in which item arrivals and demand arrivals form independent

Poisson processes. We show how to compute all performance measures of interest for this PIS. Thereafter, extensions in several directions are reviewed, among them (i) PIS’s with finite capacity and waiting demands; (ii) PIS’s with renewal item arrival times; (iii) batch arrivals of items or demands; (iv) actuarial valuation;

(v) optimization and control. Some novel contributions are also introduced.

TB15

15- West 202- CC

Software Demonstration

Invited Session

1 - Frontline Systems, Inc. - Analytic Solver Platform:

No-Compromise Analytics in Microsoft Excel

Daniel Fylstra, President, Frontline Systems Inc., P.O. Box 4288,

Incline Village NV 89450, United States of America, [email protected]

Analytic Solver Platform, a deep integration of our Premium Solver, Risk Solver and XLMiner software, is a complete toolkit for predictive and prescriptive analytics – data mining, optimization and simulation. With a far lower learning curve than other analytics software, and unmatched performance on challenging large-scale models, it offers analysts “extreme productivity.”

2 - Maximal Software - Deploying MPL Optimization Models on

Tablet Computers and Mobile Platforms

Bjarni Kristjansson, President, Maximal Software, Inc.,

933 N. Kenmore Street, Suite 218, Arlington VA 22201,

United States of America, [email protected]

The IT industry is currently undergoing a major shift to new platforms such as tablet computers and mobile phones. We will demonstrate a new server-based version of MPL OptiMax that makes writing mobile applications a relatively quick and easy process. We will take you through all the steps of implementing optimization projects - finally deploying the project on a server for servicing both web and mobile clients, using standard programming languages, such as C/C++ or

Python.

TB17

17- West 208 B- CC

Joint Session INFORM-ED/CPMS:Teaching OR/MS

Related Internships and Independent Studies

Sponsor: INFORM-ED & CPMS, The Practice Section

Sponsored Session

Chair: Eric Huggins, Associate Professor of Management,

Fort Lewis College, Durango, CO, 81301, United States of America, [email protected]

1 - Applying Big OR/MS Concepts at a Small College in a Small

Town

Eric Huggins, Associate Professor of Management, Fort Lewis

College, Durango, CO, 81301, United States of America, [email protected]

Managing OR/MS related student internships in a small town creates unique challenges and opportunities. We discuss these challenges as well as successful strategies to develop and implement internships locally. We focus on one very successful internship, but draw insights from several others that had varying degrees of success.

2 - Three Keys to Successful Research Projects:

Training, Teamwork, Timing

Mike Veatch, Gordon College, Department of Mathematics, 255

Grapevine Rd, Wenham, MA, 01984, United States of America,

[email protected]

Our experience is that undergraduates can make significant contributions to O.R.

research projects, spanning from software and numerical tests to model formulation, bounds, and derivations. Success hinges on picking highly motivated students, a carefully focused training program, building a project team, and creating sufficient time for student and advisor.

3 - A Taste of the Real World: Senior Capstone Projects

Susan Martonosi, Harvey Mudd College, Department of

Mathematics, Claremont, CA, United States of America, [email protected]

A senior capstone is a project conducted during a student’s final year, whose intention is to synthesize material learned in the classroom. For 50 years, the

Harvey Mudd College Clinic Program has been a capstone experience in which teams of undergraduates work on a real-world problem of interest to a sponsoring organization. Over 1400 projects have been completed, many of them related to operations research. This talk will provide an overview of the program and suggest best practices.

290

INFORMS Phoenix – 2012

TB18

18- West 208 A- CC

Scheduling and Logistics

Sponsor: Optimization/Computational Optimization and Software

Sponsored Session

Chair: R. Logen Logendran, Professor, Oregon State University, School of Mech, Indust, and Mfgr Engr., 204 Rogers Hall, Corvallis, OR, 97331-

6001, United States of America, [email protected]

1 - Decomposition Based Lower Bounds for a Dynamic

PCB Assembly System

Mohammad Yazdani Sabouni, PhD Student, Oregon State

University, School of Mech, Indust, and Mfgr Engr., 204 Rogers

Hall, Corvallis, OR, 97331-6001, United States of America, [email protected], R. Logen Logendran

We introduce the problem of integrating internal (machine) and external (kitting) setup times and a dynamic assembly system in the assembly of printed circuit boards (PCBs). Two algorithms are developed for this problem. Since obtaining the optimal solutions is practically difficult or even impossible, to evaluate the effectiveness of these algorithms, we use the Branch-and-Price methodology to decompose the original problem and iteratively construct a lower bound.

2 - Sequence-Dependent Group Scheduling in Hybrid Flowshops

Mir Abbas Bozorgirad, PhD Student, Oregon State University,

School of Mech, Indust, and Mfgr Engr., 204 Rogers Hall,

Corvallis, OR, 97331-6001, United States of America, [email protected], R. Logen Logendran

We develop tabu search based algorithms to address a bi-criteria group scheduling problem in hybrid flowshops with the objective of minimizing the weighted sum of total weighted completion time and total weighted tardiness. There is a sequence-dependent setup time between different groups, and the job release times and machine availability times are considered to be dynamic. The performances of algorithms are evaluated and compared to each other to find the best algorithm for this problem.

3 - Inventory Deployment Decisions for Blockbuster’s Online Movie

Rental Business

Kyung Sung Jung, PhD Candidate, University of Texas at Dallas,

Jindal School of Management, 800 W Campbell Rd., SM30,

Richardson, TX, 75083, United States of America, [email protected], Casey Chung, Shun-Chen Niu,

Chelliah Sriskandarajah

Blockbuster has a subscriber-based online movie rental industry. Subscribers maintain rental queues and their demands are fulfilled corresponding to the higher queue positions. We model the demand of DVDs by using Blockbuster datasets, determine the initial order quantity, and propose the method to allocate

DVDs to its distribution centers.

4 - Non-Permutation Schedules in Hybrid Flowshops: Tabu Search

Algorithm Versus Genetic Algorithm

Mir Abbas Bozorgirad, PhD Student, Oregon State University,

School of Mech, Indust, and Mfgr Engr., 204 Rogers Hall, Corvallis,

OR, 97331-6001, United States of America, [email protected], R. Logen Logendran

In a non-permutation schedule, positions of jobs are correlated, i.e. the position of a job in a stage can affect its positions in other stages. Tabu search (TS) is very effective in dealing with scheduling problems, but it only allows single perturbations in generating new solutions, so it may not capture those interdependencies. Therefore, TS is compared to genetic algorithm, which performs multiple perturbations, to find the best algorithm in hybrid flowshops with a bi-criteria objective.

TB19

19- West 211 A- CC

Disease Prevention and Immunization

Contributed Session

Chair: Brandon Pope, Purdue University, 315 N. Grant Street,

West Lafayette, IN, United States of America, [email protected]

1 - What Is the Right Balance between Prevention and

Treatment of Disease?

George Miller, Institute Fellow, Altarum Institute, 3520 Green

Court, Suite 300, Ann Arbor, MI, 48105, United States of America, [email protected]

There is little quantitative substantiation of the common assertion that disease prevention receives inadequate funding compared to treatment. We address this seeming paradox with a review of the arguments and the underlying evidence.

We assess the amount spent on prevention, summarize the relative cost-

TB20

effectiveness of prevention and treatment, analyze the interacting impacts of prevention and treatment spending levels on morbidity and mortality, and summarize research needs.

2 - Linking Disease and Supply Chain Models for

Immunization Programs

Sheng-I Chen, University of Pittsburgh, 1033 Benedum Hall,

Pittsburgh, PA, 152261, United States of America, [email protected],

Bryan A. Norman, Jayant Rajgopal

Many disease transmission models have been applied to help make decisions about infectious disease control. Knowing disease transmission dynamics can impact vaccine distribution strategies. We propose a mechanism for linking disease transmission and vaccine distribution models and consider how supply chain constraints affect the health benefits of immunizations programs.

3 - Management of Chronic Illnesses through Prevention on US

Healthcare Delivery System

Sameer Kumar, Qwest Endowed Chair and Professor of Operations and Supply Chain Management, University of St. Thomas,

1000 LaSalle Avenue, Minneapolis, MN, 55403,

United States of America, [email protected]

A detailed closed loop business framework for system modeling on chronic disease prevention is proposed to study the interactions among two chronic diseases, related health risk factors, using regression analysis, preventive programs and associaed costs.Optimization provides minimum threshold values of prevention programs costs and reduction effect on health risk factors to achieve breakeven values for costs of prevention equal to savings from reduced direct treatment costs over time.

4 - Information-Based Incentives: Preventive Behaviors for

Coronary Heart Disease

Brandon Pope, Purdue University, 315 N. Grant Street, West

Lafayette, IN, United States of America, [email protected],

Abhijit Deshmukh, Andrew Johnson, James Rohack

Incentives are increasingly promoted to support distributed decision-making in healthcare systems, including preventive behaviors. While the incentives literature focuses on financial reform, incentives are generally any mechanism that affects decision-making. We use a learning process of consumer behavior to model a policy maker’s information provision problem as a MDP, and solve for optimal information provision policies regarding preventive behaviors pertinent to coronary heart disease.

5 - A MDP Model for Developing Breast and Ovarian Cancer

Intervention Strategies for BRCA1/2 Carriers

Mehrnaz Abdollahian, PhD Student, Department of Industrial and

Management Systems Engineering, University of South Florida,

2038 Gregory Drive, Tampa, FL, 33613, United States of America, [email protected], Tapas Das

More than 300,000 women in the United States have the BRCA1/2 genes. They inherit 5 to 20-fold increased risks of developing breast and ovarian cancers. Costeffective intervention strategies while decreasing the risks may save up to $800 million spent each year on treatment. We develop a Markov Decision Process model and use it to find an optimal intervention strategy based on data available in the literature on the various stages of the breast and ovarian cancer and the costs of interventions.

TB20

20- West 211 B- CC

Data for Health Care

Contributed Session

Chair: Keith Willoughby, University of Saskatchewan, 25 Campus

Drive, Saskatoon, SK, S7N 5A7, Canada, [email protected]

1 - Essential Data Resources, Strategies and Research Efforts to

Inform Healthcare Reform Initiatives

Steven Cohen, Director, CFACT, AHRQ, 540 Gaither Road,

Rockville, MD, 20850, United States of America, [email protected]

Policymakers depend on model-based estimates of the future state under alternative demographic, economic and technological assumptions to complement current state assessments. They estimate the impact of changes in financing, coverage and reimbursement policy, determining who benefits and who bears the cost. This presentation focuses on data capacity, statistical quality,research and modeling efforts to inform health reform initiatives with attention to the Medical

Expenditure Panel Survey.

291

TB21

2 - A System Model for Implementing and Sustaining Evidencebased Healthcare Practice

Caitlin Hawkins, University of Southern California,

3715 S McClintock Ave., GER 240, Los Angeles, CA,

United States of America, [email protected], Shinyi Wu

There is a need to develop a model to guide sustainable implementation of evidence-based healthcare practice at the local level. Through analyzing successful implementation and existing frameworks, a four-component model is proposed comprising data, behavior, support system, and culture. The model defines the context of an implementation by combining approaches from human factors engineering, organizational theory, and systems engineering. The model is tested against a national implementation.

3 - A Multi-agent System Model to Study US Health System

Pai Liu, University of Southern California, 2355, Scarff St., Apt 1,

Los Angeles, CA, 90007, United States of America, [email protected], Shinyi Wu

We proposed a multi-agent system model to study the US health delivery system and policy making. Our model includes interacting agents representing the key players in the health system. The agents would respond to the policy by changing their behaviors. This model can serve as a tool for decision makers to test different policy designs and scenarios to understand the complexity and mitigate the risks involved.

4 - Economic Analysis of the Hospitalist Program in the Saskatoon

Health Region

Keith Willoughby, University of Saskatchewan, 25 Campus Drive,

Saskatoon, SK, S7N 5A7, Canada, [email protected],

Tim Kent, George Tannous

Recently, the Saskatoon Health Region launched a hospitalist program. Under this program, physicians work out of the hospital and primarily care for inpatients.

We examined the hospitalist program’s impacts on inpatient length of stay, readmissions, and rate of mortality. Although our results do not provide overwhelming evidence in support of the hospitalist program, we did identify a reduction in length of stay due to a change in physician payment structure.

TB21

21- West 212 A- CC

INFORMS Phoenix – 2012

Security Games and Optimization on Networks

Sponsor: Optimization/Networks

Sponsored Session

Chair: Yevgeniy Vorobeychik, Sandia National Labs,

7011 East Ave, Livermore, CA, 94550, United States of America, [email protected]

1 - Adversarial Patrolling on Networks

Yevgeniy Vorobeychik, Sandia National Labs,

7011 East Ave., Livermore, CA, 94550, United States of America, [email protected]

We will address the problem of patrolling physical infrastructure on a network when an adversary aims to inflict damage on a target. We show that, in general, there exists a Stackelberg equilibrium in Markovian stationary strategies, and present general-purpose solution techniques. Additionally, we offer highly scalable methods for special cases.

2 - Cutting Plane Algorithms for Robust Multicommodity

Network Design

Richard Chen, Senior Member of Technical Staff, Sandia National

Laboratories, P.O. Box 969 MS 9155, Livermore, CA, 94551,

United States of America, [email protected], Robert Carr,

Ojas Parekh, Cynthia Phillips

We present algorithms for solving the robust multicommodity network design problem. Given point-to-point demands, specified robustness requirements characterized by a budget-constrained disruption set and a capacity for each edge, find the minimum cost capacity expansion satisfying the given demands. In this talk I will describe the underlying problem, the model and the main components of our algorithms. Computational results from the Survivable Network Design

Library are reported.

3 - Double Oracle Algorithms for Game-Theoretic

Resource Allocation

Milind Tambe, University of Southern California, Los Angeles, CA,

United States of America, [email protected], Manish Jain, Jason Tsai

Many strategic actions carry a component beyond the immediate locale of the effort itself. To address the exponentially large strategy spaces that result, the

RUGGED algorithm for road network domains uses best-response oracles and scales to real road networks of southern Mumbai. However, when contagion is present, computing best-responses is intractable. Thus, we extend RUGGED using heuristic oracles that enable it to scale to realistically sized leadership networks from counterinsurgency.

4 - Computing Optimal Strategies to Commit to in Stochastic

Games

Joshua Letchford, Duke University, 308 Research Drive, Campus

Box 90129, Durham, 27708-0129, United States of America, [email protected], Liam MacDermed, Vincent Conitzer, Ronald Parr,

Charles Isbell

Significant progress has been made recently in the following two lines of research in the intersection of AI and game theory: (1) the computation of optimal strategies to commit to (Stackelberg strategies), and (2) the computation of correlated equilibria of stochastic games. We unite these two lines of research by studying both theoretically and experimentally the computation of Stackelberg strategies in stochastic games.

TB22

22- West 212 B- CC

Branching in Mixed-Integer Programming I

Sponsor: Computing Society

Sponsored Session

Chair: John Chinneck, Carleton University, Systems and Computer

Engineering, Ottawa, On, K1S 5B6, Canada, [email protected]

1 - The Watermelon Algorithm for the Bilevel Integer Linear

Program

Lizhi Wang, Iowa State University, 3016 Black Engineering, Ames,

IA, 50011, United States of America, [email protected], Pan Xu

Discrete bilevel optimization problems have been studied for more than two decades, but there are only a handful of algorithms, most of which are heuristic in nature. We present a so-called watermelon Algorithm for the bilevel integer linear programming problem, which can be proved to terminate finitely to the global optimal solution, if one exists. Computational experiment results will be reported to demonstrate the effectiveness of the algorithm.

2 - Branching on General Disjunctions

Ted Ralphs, Associate Professor, Lehigh University, 200 West

Packer Avenue, Bethlehem, PA, 18015, United States of America, [email protected], Serdar Yildiz

The use of general disjunctions for branching has been experimentally shown to decrease the size of the search tree. However, selecting good disjunctions is expensive and can easily lead to slower running times overall. In this talk, we focus on recent experimental work on choosing branching disjunctions heuristically. The methodology exploits the close connection between disjunctions used for branching and those used for cutting.

3 - Branching Objects in the CPLEX MIP Optimizer

Ed Klotz, Mathematical Programming Specialist, IBM,

926 Incline Way, Suite 100, Incline Village, NV, 89451,

United States of America, [email protected]

While usually described in terms of integer variables, the branch and bound algorithm can branch on other elements of a MIP as well. This presentation will consider the strengths and weaknesses of additional branching objects supported by IBM’s CPLEX optimizer, including special ordered sets, indicator constraints and general disjunctions. It will then examine some models where general disjunctions dramatically improve performance compared to branching on integer variables.

4 - Fast MILP Feasibility using General Disjunctions

John Chinneck, Carleton University, Systems and Computer

Engineering, Ottawa, On, K1S 5B6, Canada, [email protected], Hanan Mahmoud

Branch and bound algorithms for MILP branch on a single variable at each node.

General linear expressions of multiple variables are not often used because it is difficult to quickly find an effective general disjunction, and the overall algorithm may slow considerably. We show how to (i) trigger a general disjunction only when it is beneficial, and (ii) construct effective general disjunctions very quickly.

We show performance improvements against a state of the art commercial MILP solver.

292

INFORMS Phoenix – 2012

TB23

23- West 212 C- CC

Innovation in Analytics Award: Semi-finalist

Presentations II

Sponsor: Analytics

Sponsored Session

Chair: Michael Gorman, University of Dayton, Dayton, OH,

United States of America, [email protected]

1 - Using Shapley Value Analytics to Set Customer Satisfaction

Priorities Based on Kano Theory

Kenneth M. Powaga, Sr. Vice President, GfK Custom Research,

LLC, 8401 Golden Valley Road, Minneapolis, MN, 55427,

United States of America, [email protected], Stan Lipovetsky,

W. Michael Conklin

A key challenge with Customer Satisfaction Measurement is modeling the relationship between outcome and performance measures to deliver results that are actionable. This presentation demonstrates a unique method of Key Driver

Analysis, KDA, based on Shapley Value Analysis from Cooperative Game Theory.

It fits the non-linear and non-additive Kano Customer Satisfaction model by separating key drivers of dissatisfaction from the drivers of customer delight. KDA is an original non-regression technique which has been implemented with a webtool and used globally for many clients in a variety of industries with actionable results that are validated with increased customer loyalty.

2 - Fully Adaptive Designs for Clinical Trials: Simultaneous

Learning from Multiple Patients

Vishal Ahuja, PhD Candidate, University of Chicago, Booth

Business School, Chicago, IL, 60637, United States of America, [email protected], John Birge

Traditional clinical trials are randomized and the goal is to maximize learning about treatment. Adaptive designs, on the other hand, allow clinicians to learn about treatment effectiveness during the course of the trial. In an adaptive design, allocation of patients to treatments evolves dynamically as the trial progresses. An ideal adaptive design maximizes patient health outcomes without sacrificing any potential learning. We propose such a design, one that fully exploits learning from multiple patients simultaneously. We demonstrate our design’s effectiveness on a recent trial. Our design is general and applicable to any Markov decision process setting where learning takes place.

3 - Whole-hospital Operational Forecasting System

David S. Toledano, GE Global Research Center, Niskayuna, NY,

12309, United States of America, [email protected], Onur I.

Dulgeroglu, Bex G. Thomas, Kunter S. Akbay, Peter L. Katlic

Hospitals make complex patient flow decisions throughout the day to manage admissions from Emergency, Surgery and elsewhere into a scarce supply of inpatient beds, with relatively little decision support to identify those likely to be over-subscribed. In partnership with the Mount Sinai Medical Center (MSMC) in

New York, we are prototyping a novel whole-hospital forecasting system which accurately estimates patient flow movements and bed shortages over the next 24 hours and displays operational feedback through interactive dashboards. Our system utilizes real-time data and historical patterns within a discrete-event simulation and currently achieves in excess of 98% accuracy on forecast census at the hospital level.

TB24

24- West 213 A- CC

Analyzing Policy Options

Sponsor: Health Applications Society

Sponsored Session

Chair: Saied Samiedaluie, McGill University, 1001 Sherbrooke West,

Montreal, QC, H3A1G5, Canada, [email protected]

1 - The Impact of Private Providers and Cross-border Policy on

Public Healthcare Services

Dimitrios Andritsos, HEC Paris, 1 Rue de la Liberation, Paris,

France, [email protected], Christopher Tang

We examine the impact of competition from private hospitals and a recent crossborder healthcare policy on public healthcare systems in the EU. We find that in border regions, where the cost of crossing the border is low, “outsourcing” the high-cost country’s elective care services to the low-cost country is a viable strategy from which both countries’ systems can benefit.

2 - The Treatment Trap: Modeling Physicians’ Prescribing Behavior

Tinglong Dai, PhD Candidate, Tepper School of Business, Carnegie

Mellon University, 5000 Forbes Ave., Pittsburgh, PA, 15217,

United States of America, [email protected], Sridhar Tayur, Mustafa

Akan

TB25

The Congressional Budget Office estimated that $700 billion per year is spent on unnecessary procedures. In the field of cardiology, above 25% of stent procedures, which are costly and can cause complications and infections, are unnecessarily prescribed. We develop a theoretical framework of physicians’ prescribing behavior under congestion. For both inpatient and outpatient settings, we show a few operational and policy insights that depart from conventional wisdom.

3 - Optimal Inpatient Discharge Timing: A Simulation Study

Nan Kong, Assistant Professor, Purdue University, 206 S. Martin

Jischke Drive, West Lafayette, IN, United States of America, [email protected], Pratik Parikh, Charuhas Thakar,

Elizabeth Crawford

We focus on inpatient discharge timing, an important decision in acute hospital management. Academic literature and hospital practice indicate that medicallyunnecessary delays in discharging patients could lead to a reduction in effective resource capacity and ED crowding, while aggressive discharge may result in higher readmissions and increased system load. In a simulation study, we analyze discharge policies considering resource utilization and patient wait times as key performance measures.

4 - Admission Policies in a Neurological Hospital Ward

Saied Samiedaluie, McGill University,

1001 Sherbrooke West, Montreal, QC, H3A1G5, Canada, [email protected], Beste Kucukyazici, Dan Zhang,

Vedat Verter

In this paper, we study patient admission policies in a neurological hospital ward, where there are multiple patient types with different medical characteristics. The patients need to wait in ED or ICU until a hospital bed is assigned to them. Each type of patient has different arrival rate, average length of stay and waiting cost.

The problem is formulated as an average cost dynamic program over infinite horizon. An approximation scheme to solve the dynamic programming will be presented.

TB25

25- West 213 B- CC

Healthcare Analytics

Sponsor: Health Applications Society

Sponsored Session

Chair: Bruce Golden, University of Maryland, Robert H. Smith School of Business, College Park, MD, United States of America, [email protected]

Co-Chair: Edward Wasil, Kogod School of Business, American

University, Washington, DC, United States of America, [email protected]

1 - Does Multitasking Improve Productivity? Evidence from the

Emergency Department

Diwas KC, Emory University, 1300 Clifton Road, Atlanta, GA,

30322, United States of America, [email protected]

We examine the effect of multitasking on performance in an emergency department. By drawing on recent findings in the experimental psychology literature we develop several hypotheses for the effect on performance. We find that multitasking has implications for the service encounter, including patient flow and quality of care.

2 - Data-driven Appointment Scheduling in the Presence of No-shows

Michele Samorani, Alberta School of Business, University of

Alberta, Edmonton, AB, Canada, [email protected],

Linda LaGanga

We solve the problem of scheduling outpatient appointments in the presence of no-shows. Predictive analytics is used to forecast the show outcome of appointment requests, which, based on this prediction, are optimally scheduled.

Descriptive analytics is used to interpret the output of this optimization problem in order to derive a heuristic policy. The prediction quality determines whether a clinic should adopt a single-day scheduling horizon or not, and whether it should overbook or not.

3 - Data Mining to Aid Beam Angle Selection for Intensitymodulated Radiation Therapy

Stuart Price, Robert H. Smith School of Business, University of

Maryland, College Park, MD, United States of America, [email protected], Hao Howard Zhang, Bruce Golden

A good IMRT plan is a set of beam angles and intensities that deliver radiation to the tumor while minimizing damage to healthy tissue. Fully evaluating intensities and performance from angle sets is computationally intensive, we data mine evaluated plans to create a model to screen angle sets before evaluation.

293

TB26

INFORMS Phoenix – 2012

4 - Early Detection of Bioterrorism: A Combined Social and Spatial

Network Analysis

Inbal Yahav, Assistant Professor, Bar Ilan University,

Graduate School of Business, Ramat Gan, 52900, Israel, [email protected], Sean Barnes, Bruce Golden, Edward Wasil

Bioterrorism is a significant threat to the US. The spread of certain agents may be difficult to distinguish from a seasonal flu outbreak. We simulate epidemic and bioterrorism outbreaks and propose a method for differentiating between the two scenarios when only a small proportion of the population becomes infected.

TB26

26- North 221 A- CC

Operational Models of Mergers and Alliances

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Soo-Haeng Cho, Carnegie Mellon University, 5000 Forbes

Avenue, Pittsburgh, PA, 15213, United States of America, [email protected]

1 - Strategic Supplier Alliances under Default Risk

Xiao Huang, Assistant Professor, Concordia University,

1455 de Maisonneuve Blvd. W, Montreal, QC, H3G1M8, Canada, [email protected], Tamer Boyaci, Mehmet Gumus,

Saibal Ray, Dan Zhang

We study alliance formation among a set of complementary/substitutable suppliers selling to an assembler/buyer. The suppliers balance the strategic benefit of joining large alliances (lower chance to default) vs operational benefit to stay with small alliances (higher profit allocation). Coalition-proof stable alliance structures are characterized. We also find that the assembler and the buyer’s investment strategy in upstream risk construction have contrasting impact on the supplier community.

2 - United We Stand? Coordinating Capacity Investments in

Joint Ventures

Guillaume Roels, Assistant Professor, UCLA, 110 Westwood Plaza,

Los Angeles, 90095, United States of America, [email protected], Philippe Chevalier, Ying Wei

In this talk we study how to structure a joint venture between manufacturing firms that pool their resources in order to reduce their overall demand risk. In particular, we study whether capacity should be owned jointly or separately so as to ensure the joint-venture’s long-term viability, depending on the number of partners, the magnitude of economies of scale, and the degree of asymmetry between partners.

3 - Vertical Integration under Competition: Forward, Backward, or No Integration?

Ali Parlakturk, Assistant Professor, University of North Carolina-

Chapel Hill, Kenan-Flagler Business School, Chapel Hill, NC,

27599, United States of America, [email protected],

Yen-Ting Lin, Jayashankar M. Swaminathan

We consider two competing supply chains, each consisting of a supplier, a manufacturer and a retailer. The supplier controls the product quality, and the retailer sets the retail price. Each manufacturer can choose to (1) forward integrate, (2) backward integrate, or (3) staying disintegrated. We analyze the resulting equilibrium supply chain structure, and examine the effects of competition, demand perishability, and cost of improving quality on manufacturers’ choice.

4 - Newsvendor Mergers

Xin Wang, Carnegie Mellon University, 5000 Forbes Avenue,

Pittsburgh, PA, 15213, United States of America, [email protected], Soo-Haeng Cho

We study how a merger of two price-setting competitive newsvendors will affect the prices, inventories, and expected profits of merging newsvendors and their competitors in an oligopolistic market. By evaluating the pooling effect of a merger and stock-out competition under uncertain demand, we demonstrate that demand uncertainty, which has been ignored in the economics literature, can have a significant influence on both direction and magnitude of the impact of a merger.

TB27

27- North 221 B- CC

Modeling and Learning Demand

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Adam Mersereau, University of North Carolina-Chapel Hill, CB

3490, Chapel Hill, NC, 27599, United States of America, [email protected]

1 - Dynamic Pricing Model for a Duopoly Market

Xin Liu, Postdoc, University of Minnesota, Institute for

Mathematics and it’s Applications, Minneapolis, MN, 55455,

United States of America, [email protected], William Cooper

We study a duopoly with two sellers. Each of the sellers has one product. We focus on a linear demand model, namely, the demand of each product depends on prices of both products linearly. We obtain a sufficient condition, under which the prices converges to the Nash equilibrium prices. Such a sufficient condition can be formulated in terms of the covariance matrix of the prices. In particular, the correlation between two prices play an important role for the asymptotic optimality.

2 - On Information Distortion in Online Ratings

Omar Besbes, Columbia Business School, 3022 Broadway, New

York, NY, 10027, United States of America, [email protected],

Marco Scarsini

We analyze the interplay between past and future ratings due to the sequential nature of reports. We show that, for a broad class of behavioral models, and mixture thereof, the long-run average of sequentially declared ratings, when well defined, preserves stochastic order of the true rating. In addition, we show that there exist behaviors that allow manipulations of the long-run average of sequentially declared ratings through a finite number of fake reviews.

3 - Promotion Optimization for Grocery Retailers

Georgia Perakis, Massachusetts Institute of Technology, Cambridge,

MA, United States of America, [email protected], Maxime

Cohen, Zachary Leung, Kiran Panchamgam, Andrew Vakhutinsky

Promotion planning is an important process for retail grocery chains. We formulate the problem of optimizing promotions for products in a category, i.e.

deciding which products to promote, using which promotion vehicles, and at what prices. Our demand model captures several key effects that arise in practice

(complementary and substitution effects, promotion fatigue). This gives rise to a hard problem. We propose and compare various solution methods for the problem.

4 - Demand Estimation from Censored Observations with Inventory

Record Inaccuracy

Adam Mersereau, University of North Carolina-Chapel Hill,

CB 3490, Chapel Hill, NC, 27599, United States of America, [email protected]

We consider a newsvendor estimating a demand distribution from historical sales data. The newsvendor uses stocking levels to account for censoring but these stocking levels are subject to unobserved errors. We discover a systematic downward bias in demand estimates even while service levels may appear to the firm to exceed targets. We propose a heuristic correction that sharply reduces the bias.

TB28

28- North 221 C- CC

Sourcing Strategies for Retailers and Manufacturers

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Damian Beil, Associate Professor, Stephen M. Ross School of

Business, University of Michigan, 701 Tappan St, Ann Arbor, MI,

48103, United States of America, [email protected]

1 - Bargaining for an Assortment

Sebastian Heese, Associate Professor, Indiana University

Bloomington, 1309 E. 10th Street, Bloomington, IN, 47405,

United States of America, [email protected], Goker Aydin

Consider a retailer who must compose an assortment from products offered by different manufacturers. We propose a model in which the retailer engages in simultaneous bilateral negotiations with individual manufacturers. We characterize the equilibrium assortment and profit allocation, and we explore how the equilibrium depends on manufacturer and product characteristics.

294

INFORMS Phoenix – 2012

TB30

2 - Framework Agreements in Procurement: An Auction Model and

Design Recommendations

Gabriel Weintraub, Columbia Business School, 3022 Broadway,

New York, NY, 10027, United States of America, [email protected], Yonatan Gur

Framework agreements (FAs) are procurement mechanisms that are commonly used to satisfy demand that arises randomly over time. We develop the first model of suppliers’ risks in FAs, focusing on cost uncertainty. We study different mechanisms that mitigate these risks and reduce buying prices. Our results provide important prescriptions that are being used by the Chilean government to improve the design of their FAs worth a billion dollars per year.

3 - Deploying Test Auctions to Assist with Supplier Qualification

Decision-Making

Brendan See, PhD Candidate, University of Michigan,

1205 Beal Ave., Ann Arbor, MI, 48104, United States of America, [email protected], Damian Beil, Izak Duenyas

Entrant suppliers must undergo a costly qualification process prior to competing in a procurement auction. We evaluate when a buyer can benefit from holding multiple auctions when the supply base consists of both qualified incumbents and not-yet-qualified entrant suppliers who can be qualified at a cost. The buyer faces a trade-off between increased supplier competition and revealing supplier cost information. We extend the model to incorporate a credible reserve price and

TIOLI offer.

4 - Demand Estimation and Ordering under Censoring: Stock-out

Timing is (almost) All You Need

Tong Wang, Assistant Professor, National University of Singapore,

15 Kent Ridge Drive, Singapore, 119245, Singapore, [email protected], Aditya Jain, Nils Rudi

Retailers learn about demand by observing sales. However, this learning is limited by the amount of inventory a retailer carries. This loss of information due to censoring requires a retailer to carry excess inventory. We propose a new way of learning about demand with significantly simple sufficient statistics. We analyze this demand learning method for a simple multi-period Newsvendor setting. We provide a comparison between stock-out timing and stock-out event based demand updating.

TB29

29- North 222 A- CC

Empirical Investigations into Delivery of Healthcare

Sponsor: Manufacturing & Service Oper Mgmt/ Healthcare

Operations/SIG

Sponsored Session

Chair: Sarang Deo, Assistant Professor, Indian School of Business,

Gachibowli, Hyderabad, 500032, India, [email protected]

1 - Stress on the Ward: Evidence of Safety Tipping in Hospitals

Stefan Scholtes, University of Cambridge, Judge Business School,

Cambridge, CB2 1AG, United Kingdom, [email protected],

Ludwig Kuntz, Roman Mennicken

We confirm the existence of workload-related safety tipping points for in-hospital mortality using discharge records of 83,485 patients in 234 clinical departments.

We use a survival model with discharge selection and estimate that 12% of all deaths of patients who experience workload above the tipping point can be explained by workload.

2 - Docs under Load: The Endogenous Nature of Work Content in an Emergency Department

Robert Batt, The Wharton School, University of Pennsylvania,

3730 Walnut St., Huntsman Hall, Suite 500, Philadelphia, PA,

19104, United States of America, [email protected],

Christian Terwiesch, Olan Soremekun

We present an empirical study of an emergency department that focuses on the endogenous nature of work content per patient. We show that triage nurses & physicians adjust the number of diagnostic tests ordered per patient in response to the level of crowding in the system. During times of crowding, more diagnostic testing is ordered at triage in an effort to improve throughput. However, this leads to an overall increase in testing when the system is most congested.

3 - Analytics for OR Access at a Large Teaching Hospital

Retsef Levi, Massachusetts Institute of Technology,

100 Main Street, Building E62-562, Cambridge, MA,

United States of America, [email protected], Vivek Farias,

Ryan Graue

We leverage predicability in seemingly highly variable utilization patterns via tools that (a) predict utilization of a user defined set of surgical resources at some future time using relevant historical patterns and existing block booking data and

(b) consolidate resource utilization in a manner that respects surgeon preferences/ constraints. These tools help Beth Israel Deaconess Medical Center identify and utilize surgical resources that would otherwise have been lost.

295

4 - Heterogeneity in Physician Performance: Efficiency vs. Flexibility

Sarang Deo, Assistant Professor, Indian School of Business,

Gachibowli, Hyderabad, 500032, India, [email protected],

Aditya Jain

We use patient flow data at a not-for-profit, specialty eye care clinic to investigate the impact of case-mix heterogeneity at the physician level on time spent by patients in the clinic after controlling for other operational factors such as overall workload, complexity of patients, fresh vs. follow-ups, appointments vs. walk-ins, paying vs. non-paying etc. We use the results of our analysis to empirically characterize the trade-off between efficiency and flexibility of individual physicians.

TB30

30- North 222 B- CC

Commodity Procurement and Processing

Sponsor: Manufacturing & Service Oper Mgmt/iFORM

Sponsored Session

Chair: Ankur Goel, Case Western Reserve University, Cleveland,

Cleveland, United States of America, [email protected]

1 - Procurement Strategies for Multiple Inputs

Pascale Crama, Singapore Management University, 50 Stamford

Road, Singapore, 78899, Singapore, [email protected],

Onur Boyabatli

This paper analyzes the procurement decision of a firm that produces a single commodity output using two commodity inputs. The firm can source the two inputs in advance of the selling season through long-term contracts or from the spot markets. We investigate the impact of using different contract types (fixedprice versus index-priced) on the profitability of the firm. The main motivation comes from a soda ash producer located in India that procures multiple commodity inputs for production.

2 - The Value of Operational Flexibility in Uncertain Supply and

Demand Environments

Xiaole Wu, Assistant Professor, Fudan University, Room 513,

Siyuan Building, 670 Guoshun Road, Shanghai, 200433, China, [email protected], Lingxiu Dong, Panos Kouvelis

Refining is indispensable to almost every natural resource and agricultural commodity based industry. We investigate the interconnection among input procurement, intermediate processing, and output blending decisions in the refining process. We offer insights on how the market conditions affect the value of conversion flexibility and its interaction with range flexibility.

3 - An Integrated Approach to Commodity Risk Management

Fehmi Tanrisever, Eindhoven University of Technology, Den

Dolech 2, Eindhoven, 5612, Netherlands, [email protected],

Genaro Gutierrez

In this paper, we examine the integrated operating and financial hedging decisions of a value maximizing firm, under the presence of capital market imperfections. Motivated by the flour milling industry, we consider a commodity processor who buys a certain commodity, converts it into a final product and sells it to the end customers. In presence of costly external funds, we derive the optimal value-maximizing policy and compare it with a simple short-hedge.

4 - Operational Flexibility and Valuation of Oil Refinery Operations

Sripad Devalkar, Indian School of Business, Gachibowli,

Hyderabad, 500032, India, [email protected], Ankur Goel

We model a crude oil refinery that has operational flexibility in blending different grades of crude oil and adjusting the mix of output products produced upon refining. Using a multi-period stochastic dynamic programming formulation, we explore the interconnections between the input (crude) procurement and output product mix decisions and quantify the value of blending and refining flexibilities.

TB31

INFORMS Phoenix – 2012

TB31

31- North 222 C- CC

Innovative Modern Service and Inventory Systems

Sponsor: Manufacturing & Service Oper Mgmt/Service Operations

Sponsored Session

Chair: Robert Swinney, Associate Professor, Graduate School of

Business, Stanford University, 655 Knight Way, Stanford, CA, 94305,

United States of America, [email protected]

1 - Would the Social Planner Let Bags Fly Free?

Gad Allon, Northwestern University, 2001 Sheridan Rd, Evanston,

IL, United States of America, [email protected],

Achal Bassamboo

In June 2008, American Airlines became the first major US airline to institute a fee on the first checked bag. Within 18 months, nearly every other major carrier followed suit. Here, we examine whether the fees are socially efficient. We find that the fees are in fact Pareto improving. While customers may dislike the fees, they are a means to alter customer behavior and thus reduce the airline’s costs.

Customers are compensated by lower base fares.

2 - On the Correlation between Waiting Times and Service Speed

Pnina Feldman, Haas School of Business, UC Berkeley, 545

Student Services Bldg #1900, Berkeley, CA, 94720, United States of America, [email protected], Jing Dong, Galit Yom-Tov

The service operations literature typically assumes that firms control service times and that those are independent of the time customers spend waiting. Empirical evidence suggests otherwise. In fact, the time consumers spend in line affects their expectations on service length. Some consumers require longer service after waiting a long time and others expect a rushed service. We examine how such correlations affect the firm’s operational performance.

3 - Learning Quality from Service Outcomes

Senthil Veeraraghavan, Associate Professor, The Wharton School,

3730 Walnut Street Suite 550, Philadelphia, PA, 19104, United

States of America, [email protected], Laurens Debo

We study a service provider whose service value is unknown to arriving customers. Service outcomes are random depending on the quality of the service provider. Customers decide whether to join the service, or renege from the service based on the limited service outcomes/reviews that they observe. We consider how service policies influence consumer learning and social welfare.

4 - The Groupon Effect: Threshold-Discounting with

Strategic Customers

Serguei Netessine, Professor, INSEAD, Boulevard de Constance,

Fountainbleau, 77305, France, [email protected],

Simone Marinesi, Karan Girotra

We analyze the business model of Groupon which offers discounts to customers if a sufficient demand for service is reached. We show when Groupon business model works best and suggest improvements for its design.

TB32

32- North 223- CC

Impact of Environmental Policy on the Durable

Products Supply Chain

Sponsor: Manufacturing & Service Oper Mgmt/Supply Chain

Sponsored Session

Chair: Haoying Sun, Assistant Professor, Texas A & M University,

4217 TAMU, College Station, TX, 77843, United States of America, [email protected]

1 - Design for Recovery or Obsolescence: The Impact of

Take-Back Legislation

Ximin Huang, College of Management, Georgia Institute of

Technology, Georgia Institute of Technology, 800 West Peachtree

Street NW, Atlanta, GA, 30308, United States of America, [email protected], Atalay Atasu, Beril Toktay

We consider a monopolist who has two product design options to manage the end-of-life costs/revenues associated with its products: making products more durable or recyclable. We explore how the recyclability and durability choices are affected by the requirements of take-back legislation.

2 - Effect of Government Subsidies on the Adoption of Resource

Efficient Products

Haoying Sun, Assistant Professor, Texas A & M University, 4217

TAMU, College Station, TX, 77843, United States of America, [email protected], Steve Gilbert

We use the durable goods framework to study how various forms of government subsidy programs shift consumer’s adoption of resource efficient products and how this in turn affects consumption of scarce natural resources.

3 - Regulating Markets for Valuable Waste

Gokce Esenduran, Ohio State University, College of Business,

656 Fisher Hall Fisher, Columbus, OH, 43210, United States of

America, [email protected], Luk Van Wassenhove,

Atalay Atasu

Take-back legislation mandates minimum recovery rates for waste products. The recoverable value in waste products may create competition between producer and scavenger, and divert them from landfills even under no legislation. We identify the conditions where legislator should not distort an efficient waste market by imposing recovery targets.

4 - Competing with Bandit Supply Chains

Meng Li, Phd Student, The University of Texas at Dallas,

Richardson, TX, 75080, United States of America, [email protected], Suresh Sethi, Jun Zhang

We study competition between a mainstream firm and a decentralized bandit supply chain. Due to the free-riding effect, the bandit supply chain may even offer higher quality products than the mainstream firm. Finally, the mainstream firm’s profit as a function of the free-riding effect is U-shaped, so that free-riding by the bandit supply chain may eventually benefit the mainstream firm.

TB33

33- North 224 A- CC

Inventory Management with Stochastic and

Changing Demand

Contributed Session

Chair: Burcu Aydin, HP Labs, 1501 Page Mill Rd. 1U MS 1140,

Palo Alto, CA, 94304, United States of America, [email protected]

1 - Perturbed Demand and Penalty Cost Inventory Models: Optimal

Policy Correspondence

James Lavin, NC State University, 400 Daniels Hall, Raleigh, NC,

27695, United States of America, [email protected],

Donald Warsing, Anita Vila-Parrish, Russell E. King, Semra Ahiska

In contrast to penalty cost models, perturbed demand models explicitly consider a loss of future demand as a consequence of stock-outs. In a multi-period setting, we formulate and solve a problem in which realized demand in later periods is affected by fulfillment performance in earlier periods. The financial impact of the potential revenue loss on the optimal perturbed demand policy can be directly related back to traditional penalty cost models, giving insights into the correspondence.

2 - On-line Dynamic Inventory Control with Nonstationary

Demands

Jianfeng Mao, Assistant Professor, Nanyang Technological

University, 50 Nanyang Avenue, Singapore, Singapore, [email protected]

We consider an on-line inventory control in single-echelon systems with nonstationary demands over multiple periods. Most of state-of-art inventory control methods require the assumption of i.i.d demands, which is not realistic in a disruptive supply chain. We develop a novel on-line inventory control algorithm to efficiently compute the optimal order in a real-time fashion by introducing a “best solution in probability” obtained by solving a series of off-line inventory control problems.

3 - Advance Demand Information in a Multi-Item System

Fernando Bernstein, Professor, Duke University, 100 Fuqua Drive,

Durham, 27708, United States of America, [email protected],

Greg DeCroix

We consider a firm that can obtain one of two types of advance information about demand for the two products it sells: total demand volume across products and the mix of demand between products. For each type of information, we derive closed form expressions for optimal capacities in two different settings – one with a flexible resource and one without. We then compare the impacts of the two information types and the presence/absence of flexibility on capacity choices and firm profits.

296

INFORMS Phoenix – 2012

4 - Bayesian Inventory Management with Potential

Change-points in Demand

Zhe Wang, University of North Carolina-Chapel Hill, CB 3490,

Chapel Hill, NC, 27599, United States of America, [email protected], Adam Mersereau

We consider an adaptive inventory control problem in which at some known potential change-points, the demand process may change abruptly with some probability. The underlying demand process is never revealed to the decision maker. Using a Bayesian framework, we analyze the behavior of the optimal policy as demands are observed and the decision maker’s beliefs evolve over time.

5 - Collaborative Forecasting in Inventory Management

Burcu Aydin, HP Labs, 1501 Page Mill Rd., 1U MS 1140, Palo Alto,

CA, 94304, United States of America, [email protected], J.S. Marron

We investigate implications of forecast sharing in a Collaborative Inventory

Management setting. Some of the well-known forecast drivers, together with correlations generated by the interactive nature of shared data and the inherent white noise create a rich and complex data set. We analyze the interactions of forecast values with the responses in a supply chain setting with information sharing.

TB36

organizations marketing campaigns. These models can be used to reallocate marketing resources to increase effectiveness of the marketing campaigns.

2 - Robust Revenue Management with Hybrid Information

Yingjie Lan, Peking University, Guanghua School of Management,

Beijing, 100871, China, [email protected], Michael Ball,

Itir Karaesmen

We study the joint decision of seat allocation and overbooking with hybrid information quality. Information with mixed quality arises as no-show records are accurate and more stable, while demand data is often censored and more volatile.

We optimize the worst case performance and solve the problem as a zero-sum game. Our numerical results show the practical relevance of our method.

3 - Conditional Promotions and Consumer Overspending

Thunyarat (Bam) Amornpetchkul, University of Michigan,

701 Tappan St., Ann Arbor, MI, 48109, United States of America, [email protected], Hyun-Soo Ahn, Ozge Sahin

Conditional promotions (e.g. spend $50 get 30% off, spend $50 get $15 off) are widely used in retail pricing. They are especially effective when some consumers are deal-prone (obtain transaction utility from purchasing at a discount). We compare two most common discount policies: percent-off and dollar-off, and show that both policies can lead to consumer overspending. We also show, depending on the nature of the product, one discount policy can perform better than the other.

TB34

34- North 224 B- CC

Emotions and Innovation Management

Cluster: New Product Development

Invited Session

Chair: Manuel Sosa, Associate Professor of Technology and Operations

Management, INSEAD, 1 Ayer Rajah Ave., Singapore, 138676,

Singapore, [email protected]

1 - The Design Challenges of Experiential Services

Yannis Bellos, Georgia Institute of Technology, 800 West Peachtree

St. NW, Atlanta, GA, United States of America,

[email protected], Stylianos Kavadias

In this paper we explore the design challenges of an organization that develops an experiential service. Building on the customer journey concept, which maps an experiential service to customer-provider interaction touchpoints, we analyze the service provider’s design decisions: the touchpoints she controls, and the price she charges. We fully characterize the conditions under which the provider can use her design decisions to successfully signal service quality.

2 - Idea Generation and the Role of Feedback

Joel Wooten, University of Pennsylvania, The Wharton School,

Philadelphia, PA, United States of America, [email protected], Karl Ulrich

In many innovation settings, ideas are generated over time and managers face a decision about if and how to provide in-process feedback about the quality of submissions. We use innovation tournament field experiments to examine the effect of feedback on idea generation and show individual-level differences between no feedback, random feedback, and directed feedback.

3 - A Structured Approach to Identify Styles in Designs

Tian Chan, PhD Student, INSEAD, 1 Ayer Rajah Avenue,

Singapore, 138676, Singapore, [email protected],

Jurgen Mihm, Manuel Sosa

A style is an organizing principle based on categorizing visually similar designs into a describable concept. This paper presents a structured approach to identifying styles using cluster analysis. Design patent data provides a unique opportunity to analyze style on a large scale due to a tightly controlled citation process which conveys similarity in visual impression between designs. Finally, we test styles developed using this approach against theory and perceptual experiments.

TB35

35- North 225 A- CC

New Applications in Revenue Management

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: Itir Karaesmen, American University, 4400 Mass Avenue NW,

Washington, DC, United States of America, [email protected]

1 - Donor Behavior and Nonprofit Marketing Resource Allocation

Itir Karaesmen, American University, 4400 Mass. Avenue NW,

Washington, DC, United States of America, [email protected]

We study repeat donor behavior and build statistical models to predict donor behavior based on internal and external factors associated with a nonprofit

TB36

36- North 225 B- CC

Quality, Pricing, and Channel Decisions

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: Xuying Zhao, University of Notre Dame, Notre Dame, IN,

United States of America, [email protected]

1 - Distribution Channel Structure for Competing Supply Chains with Price and Lead-Time Sensitive Demand

Lucy Gongtao Chen, NUS Business School, National University of

Singapore, 119245, Singapore, [email protected], Jihong Ou,

Zhengping Wu

We consider two competing manufacturers that sell substitutable products to customers who are sensitive to both price and delivery lead-time. Each manufacturer can sell either through a company-owned store or through an independent retailer. We identify the equilibrium channel strategies and analyze the driving forces behind the equilibrium. We also discuss how structural assumptions affect the equilibrium.

2 - Pricing Strategies in Advance Selling

Xuying Zhao, University of Notre Dame, Notre Dame, IN,

United States of America, [email protected], Zhan Pang

We study and compare three pricing strategies in advance selling: dynamic pricing, price commitment, and pre-order price guarantee. We find that a seller benefits from demand uncertainty when pre-order price guarantee is used in advance selling.

3 - Making Quality Decisions in Tough Economic Times

Leon Chu, University of Southern California, Los Angeles, CA,

United States of America, [email protected], Lian Qi,

Rachel Chen

This paper analyzes a two-dimensional model for firm’s optimal quality decisions where consumers vary both in their reserve utility for the product and in their sensitivity towards the quality. While the firm may want to offer lower quality products when the consumers’ quality sensitivities deteriorate, the optimal response towards changes in the reserve utility depends on the target consumer groups. Our study highlights the importance of separating reserve utility from the quality sensitivity.

4 - Optimal Project Size in the Presence of Moral Hazard in Teams and Limited Commitment

George Georgiadis, University of California-Los Angeles, Anderson

School of Management, 110 Westwood Plaza, Los Angeles, CA,

United States of America, [email protected], Steven Lippman,

Christopher Tang

We study the interaction between a team of agents who work to complete a project and a manager who selects its requirements. The main result is that with limited commitment, the manager extends the project as it progresses, which induces the agents to work less, thus rendering all parties worse off. We show that the manager should delegate the choice of the requirements to the agents or foster a relatively uncooperative environment within the team unless she has sufficient commitment power.

297

TB37

INFORMS Phoenix – 2012

TB37

37- North 226 A- CC

P-Median and Variants

Sponsor: Location Analysis

Sponsored Session

Chair: Erhan Kutanoglu, The University of Texas at Austin, 204 E Dean

Keeton St. C2200, Austin, TX, 78712, United States of America, [email protected]

1 - Interactive Location Decisions: An Airport Example

Yupo Chan, University of Arkansas at Little Rock, 2801 South

University, Little Rock, AR, 72204, United States of America, [email protected]

Using airport location as an example, an interactive multicriteria decision-process is discussed. Without specifying the utility/value function, the tradeoff between travel time and noise impact is accomplished using a Frank-Wolfe optimization procedure. Computational results for both network location and planar location problems are presented, which are shown to be consistent with existing theories in continuous and discrete location.

2 - An Empirical Evaluation of P-median Problem Reformulations

Roger Rios, Professor, Universidad Autonoma de Nuevo Leon,

Graduate Program in Systems Engineering, AP 111-F, Cd.

Universitaria, San Nicolas de los Garza, NL, 66450, Mexico, [email protected], Ana Laura Gonzalez

Recently, Church proposed a reformulation and solution procedure for the pmedian problem. Later, Elloumi proposed a different reformulation for the problem suggesting this could be embedded within an exact optimization method.

In this talk we present an algorithm for the p-median problem that relies in the

Elloumi’s problem reformulation. The algorithm is compared to the procedure proposed by Church on a set of benchmark instances.

3 - Economies of Scale in Hub Location Problems

James Campbell, University of Missouri-St Louis, College of

Business Administration, St Louis, MO, 63121, United States of

America, [email protected]

Basic hub location formulations use a simple model for transportation cost with a flow-independent discounted cost rate on inter-hub links to model economies of scale from flow consolidation. However, in optimal solutions inter-hub flows are often exceeded by flows on some spokes; yet the smaller inter-hub flows receive the cost discount. This paper examines optimal passenger and freight flows to document this behavior and suggests improvements in modeling economies of scale for hub location.

TB38

38- North 226 B- CC

Panel Discussion: Teaching Service Science

Sponsor: Service Science

Sponsored Session

Chair: Sheneeta White, University of St. Thomas, 2115 Summit Ave,

MCH 316, St. Paul, MN, 55016, United States of America, [email protected]

1 - Panel on Teaching Service Science

Moderator: Sheneeta White, University of St. Thomas, 2115

Summit Ave., MCH 316, St. Paul, MN, 55016, United States of

America, [email protected], Panelists: Sanjeev Bordoloi,

Turgut Akin, Jeffery Smith, Christoph Heitz

This panel will discuss all aspects of teaching service science at the undergraduate and graduate levels. The panel will compare and contrast teaching service science versus service operations. We will cover a wide array of pedagogical concerns such as textbooks, cases, course topics, assessments, etc. There will also be a discussion of the appropriateness of having a service-focused course as part of a business school curriculum.

TB39

39- North 226 C- CC

Production and Assignment Problems

Contributed Session

Chair: Fadime Uney-Yuksektepe, Assist. Prof., Istanbul Kultur

University, Atakoy Campus, Department of Industrial Engineering,

Istanbul, 34156, Turkey, [email protected]

1 - Multistage Stochastic Modelling of Global Production Planning under Uncertainty

Lin Zhu, Mr., University of Southampton, University Road,

Southampton, SO17 1BJ, United Kingdom, [email protected],

Yue Wu, Honora Smith

This paper examines a multi-period, multi-product and multi-plant global production planning problem from a global apparel manufacturing company, whose headquarters is in Hong Kong, manufacturing plants are in several Asia countries and sale departments are in Northern America and Europe. We propose a multistage stochastic recourse model to handle the uncertain market demand and import quota. Computational results demonstrate the effectiveness of the multistage model.

2 - Emergency Operations Scheduling of a Supply Chain Network

Hui Dong, Rutgers Business School, 1301 Wall Street,

Lyndhurst, NJ, 07071, United States of America, [email protected], Lei Lei, Kangbok Lee

We present a mathematical model for emergency operations scheduling of a multi-echelon supply network of medical supplies. Each demand point has an order quantity and an expected delivery time, and each facility on the network has a capacity. The objective is to minimize the total tardiness. An approximation algorithm is proposed and evaluated through simulation.

3 - A Heuristic Solution for Consultant Routing and

Assignment Problem

Yuanyuan Dong, Student, Southern Methodist University,

6425 Boaz Lane, Dallas, TX, 75205, United States of America, [email protected], Junfang Yu

This paper presents a heuristic solution for a consultant routing and assignment problem. The objective is to minimize the total consultant travel cost. Different consultants may have different skill sets and different consulting tasks may need different skills. The skill matches between the consultants and consulting tasks are considered.

4 - On Pricing and Composition of Multiple Bundles

Juan-Carlos Ferrer, Associate Professor, P. Universidad Catòlica de

Chile, Casilla 306 Correo 22, Santiago, Chile, [email protected],

Gabriel Bitran, Alejandro Cataldo

We study the problem facing a firm of determining the optimal composition and pricing of multiple bundles offered in a market with competition. We assume that the prices and characteristics of the competitor’s bundles are known and that the competition does not react in the short run to the firm’s decisions. Consumers are assumed to be rational and to maximize a random utility function. We use a mixed integer non-linear program, and a novel two-phase solution approach is therefore developed.

TB40

40- North 227 A- CC

Joint Session ENRE/SPPSN: Carbon

Abatement Policies

Sponsor: Energy, Natural Res & the Envi/ Environment and

Sustainability & Public Programs, Service and Needs

Sponsored Session

Chair: Xu (Cissy) Yang, Postdoctoral Associate, Massachusetts Institute of Technology, 1 Amherst Street, E40-211, Cambridge, MA, 02142,

United States of America, [email protected]

1 - A Carbon Emission Cost Based Economic Production

Quantity Model

Zhi Tao, Kent State University, 475 Terrace Drive, Kent, OH,

United States of America, [email protected], Alfred Guiffrida,

Felix Offodile

This paper integrates carbon emission cost into the economic production quantity model and investigates the tradeoff results of both economic concern and environmental concern. The findings provide optimal decisions on production quantity, carbon emission and minimum budget for carbon emission conscious manufacturers and also shed light on carbon emission management.

298

2 - Pollution and Production

Francois Giraud-Carrier, University of Utah, David Eccles School of

Business, 1645 E Campus Center Dr, Salt Lake City, UT, 84112,

United States of America, [email protected],

Krishnan Anand

We compare three popular pollution control mechanisms–Cap, Cap-and-Trade, and Tax– to each other and to the Groves mechanism. We compare outcomes along several economic variables, including output, firm profits and welfare. We show that firms always reduce output to meet any pollution constraint. We discuss an important implication of the output reduction effect on welfare: the tax rate that achieves Groves is higher, and the cap under Cap-and-Trade lower, than those that maximize welfare.

3 - Price Containment in Cap and Trade Policies

Yihsu Chen, Assistant Professor, University of California, Merced,

5200 N. Lake Rd, Merced, CA, 95343, United States of America, [email protected], Andrew Liu

Various instruments are available in existing or proposed policies to contain price volatility of permits. We examine the effect of a price ceiling & floor on the investment of new capacity. Besides undesirable emission consequences, we find that the price-ceiling could also lead to over-investment due to the inflated consumption induced by low power prices when the permit prices are capped. On the other hand, a permit floor would result in an under-investment in new capacity.

4 - Price vs Quantity Regulation for Price Setting Firms

Cristian Figueroa, Operations Research Center Massachusetts

Institute of Technology, 77 Massachusetts Ave. Bldg. E40-149,

Cambridge, MA, 02139, United States of America, [email protected], Xin Chen, David Simchi-Levi

In a market producing an externality requiring regulation, we analyze two methods: Pigovian tax and quotas sold for the externality. We show equivalence in a deterministic setting with full information. When demand uncertainty is introduced, we characterize equilibria. Under linear demand and quadratic costs, we show that equilibria inducing the same average price and average emissions, have different outcomes in terms of profits for the producers as well as revenue for the regulator.

5 - Estimation Model of Carbon Emissions for Freight Transport under Conditions of Megacities

Josue Velazquez, PhD Student, Monterrey Tech, Carlos Lazo 100

Santa Fe, Mexico City, Mexico, [email protected],

Edgar Blanco, Jan Fransoo

Most methodologies used to estimate carbon emissions due to freight transport activities are based on parameters of developed countries. We conduct a field study for a Parcel Company in Mexico City to provide a model to estimate carbon emissions under conditions of megacities. Using empirical data we compare the model against methodologies used in the US and Europe and the results confirm significant differences among the estimated emissions.

TB41

41- North 227 B- CC

Forestry Session III

Sponsor: Energy, Natural Res & the Environment/Forestry

Sponsored Session

Chair: Marc McDill, Associate Professor, Pennsylvania State School of

Forest Resources, 310 Forest Resources Building, University Park, PA,

16802, United States of America, [email protected]

1 - Design and Analysis of Wood Pellet Supply Chain

Mahdi Mobini, University of British Columbia, Department of

Wood Science, Faculty of Forestry, Vancouver, BC, V6T 1Z4,

Canada, [email protected], Taraneh Sowlati, Shahab Sokhansanj

We present a new simulation model for the wood pellet supply chain which considers the supply of raw materials, their transportation, production of pellets and finally distribution of pellets to domestic and international customers. It is based on a typical pellet mill in Canada and provides estimates of the cost and amount of delivered wood pellets. The model is used to evaluate different scenarios and the effects of changes in the supply chain on the outputs of the model.

2 - Estimating Future Impacts of Stochastic Wood Supply Risks

Peter Rauch, BOKU-University of Natural Resources and Life

Sciences, Institute of Production and Logistics, Feistmantelstrasse

4, 1180 Wien, Vienna, Austria, [email protected]

In Austria procurement of (combined) heating plants firing woody biomass is threatened by recent trends like climate change, restrictions on imports, booming wood demand in emerging nations, and rising bioenergy demand in neighboring countries. The System Dynamics model of the Austrian wood supply “RiskHo” was developed to examine the mid- and long-term impacts of supply risks on wood supply and includes key stochastic risk agents with Monte Carlo simulation.

INFORMS Phoenix – 2012

TB42

3 - The Computational Properties of Model 4 Harvest

Scheduling Models

Rachel St.John, University of Washington, Seattle, WA,

United States of America, [email protected], Sandor Toth

Model 4 is a new spatially explicit harvest scheduling model that uses adaptive volume coefficients, driven by a set of difference equations and Boolean algebra, to transition the states of forest stands from one planning period to the next. We provide a mathematical description of Model 4, show how maximum harvest opening size constraints can be incorporated in the new model and present some computational results

4 - Roll Assortment Planning Under Stochastic Demand and

Capacity Constraint

Satyaveer S. Chauhan, Concordia University, Montreal, QC,

Canada, [email protected], Anjali Awasthi,

Alain Martel, Sophie D’Amours

A roll can produce several paper products, however multiple parent rolls are used to balance the inventory holding and trim related costs. The problem becomes complex if processing capacity is limited and demand is stochastic. In this we present a mathematical programming based approach to tackle such problem.

TB42

42- North 227 C- CC

Urban Mobility and Land Use

Sponsor: Transportation Science & Logistics/ Urban Transportation

Sponsored Session

Chair: Yueyue Fan, Associate Professor, University of California,

Department of Civil & Environmental Eng., Davis, CA, 95616,

United States of America, [email protected]

1 - An Integer Programming Model for Controlling Urban Sprawl

Piyush Kumar, University of Texas at Arlington,

416 Yates St., Arlington, TX, United States of America, [email protected], Jay Rosenberger

Sprawl has a detrimental effect on quality of life and the environment. Ewing et al. (2002) defined measures of sprawl in the present urban structure. In this research, we develop an integer programming model to optimize land usage subject to sprawl constraints, which are based upon the sprawl measures. We describe heuristics and a Dantzig-Wolfe method to solve the model. Finally, we discuss topics of future research.

2 - Optimizing Stable Rides for Dynamic Ride-sharing Systems

Xing Wang, PhD Student, Georgia Tech, 765 Ferst Dr, Atlanta, GA,

30332, United States of America, [email protected],

Alan Erera

Smartphone technology enables dynamic ride-sharing systems. Matching drivers and riders plays a key role in solving ride-sharing problems. A system-wide optimal solution aimed at minimizing the external societal costs may not necessarily optimize the cost-savings of each individual participant. We develop optimization based algorithms to balance the system’s benefits and users’ satisfaction. We evaluate our methods by presenting a simulation study based on travel demand data from Atlanta.

3 - A Hierarchical Urban Traffic Control Problem

Paul Hoffer, University of Arizona, Tucson, AZ, United States of

America, [email protected], Yi-Chang Chiu

In this talk, we discuss a concept of hierarchical control of urban vehicular traffic.

The concept starts from partitioning the entire networks into controllable subareas. The upper-level problem aims to divert traffic among different sub-areas.

The lower problem aims to optimize local traffic within the sub-area. Various methods in network partitioning, upper- and lower-level flow control strategies are presented and discussed.

299

TB43

TB43

43- North 228 A- CC

Railway Capacity Analysis

Sponsor: Railway Applications

Sponsored Session

Chair: Matthew Petering, Assistant Professor, University of Wisconsin—

Milwaukee, P.O. Box 784, EMS E367, Milwaukee, WI, 53201,

United States of America, [email protected]

1 - Incremental Capacity Expansion – Heavier Loads, Faster

Empties and Quicker Meets

Hans Boysen, Researcher, KTH Royal Institute of Technology,

Department of Transport Science, Traffic and Logistics, Stockholm,

SE-10044, Sweden, [email protected]

To meet growing demand, a study is being done on measures to raise transportation capacity, including increased tonnage per loaded train using the existing locomotives, higher return speed for empty trains, and quicker meets.

Results show that for a trailing tonnage increase of some 19 %, the longer loading, transit and unloading times would be more than recovered by raising the empty return speed by some 10 km/h. Simultaneous entry to meets will further raise capacity and reduce cycle time.

2 - Analyzing the Progression from Single to Double Track using

Simulation Techniques

Samuel L. Sogin, Graduate Research Assistant, University of

Illinois at Urbana-Champaign, 205 N Matthews B118, Urbana, IL,

61801, United States of America, [email protected]

Much of the railway network is single track. Freight and passenger traffic is expected to increase requiring additional capacity. The funding for the complete second track may not be available all at once. This track may be phased in over time as funding becomes available. These analyses use simulation to define the operating characteristics of these hybrid track configurations. These results will facilitate the development of an optimal incremental upgrade model for capacity expansion.

3 - Mixed Integer Programming for Capacity Analysis of a Single

Track Railway Part I: Two Train Types

Matthew Petering, Assistant Professor, University of Wisconsin—

Milwaukee, P.O. Box 784, EMS E367, Milwaukee, WI, 53201,

United States of America, [email protected],

Dietrich Bergmann, Mojtaba Heydar

We analyze the capacity of a single track, unidirectional rail line. A set of intermediate stations lies between an origin and destination with one siding at each station. Two train types–express and local–are dispatched from the origin. A mixed integer program is developed to measure the capacity of the railway.

Constraints include a maximum total dwell time for the local train and headway considerations on the main line and in stations. Hundreds of problem instances are solved to optimality.

4 - Mixed Integer Programming for Capacity Analysis of a Single

Track Railway Part II: Generalized Model

Mojtaba Heydar, PhD Candidate, University of Wisconsin—

Milwaukee, P.O. Box 784, Milwaukee, WI, 53201,

United States of America, [email protected], Matthew Petering,

Dietrich Bergmann

We present a generalized mixed integer program for analyzing the capacity of a single track, unidirectional rail line. Our MIP model can handle (1) more than one siding per station, (2) more than two train types, and (3) different origins and destinations for each train type. Several heuristic and exact methods are employed in preliminary phases to properly set up the model, allowing CPLEX

11.2 to find optimal solutions to hundreds of large problem instances in a surprisingly short time.

INFORMS Phoenix – 2012

TB44

44- North 228 B- CC

Supply Chain: Managing Disruptions

Contributed Session

Chair: Paul Cronin, University of Texas at Austin, 2110 Speedway Stop

B6500, CBA 5.202, Austin, TX, 78712-1750, United States of America, [email protected]

1 - Retailer Competition under Supply Uncertainty

Shaoxuan Liu, Associate Professor, Antai College of Economics and

Management, Shanghai Jiao Tong University, 535 Fa Hua Zhen

Rd., Shanghai, China, [email protected], Rick So, Fuqiang Zhang

We use a game-theoretical framework to study the impact of supply uncertainty on the joint marketing and inventory decisions between two competing retailers.

We derive some analytical results that provide managerial insights on the equilibrium behavior of these joint decisions and the corresponding profitability of the two retailers.

2 - Effect of Process Flexibility in Mitigating Supply

Chain Disruptions

Nezir Aydin, PhD Candidate, Wayne State University,

4815 Fourth Street, Detroit, MI, 48202, United States of America, [email protected], Alper Murat

We study the value of process flexibility in mitigating supply chain disruptions. A location-allocation problem is considered where potential facilities are subject to failure and facility-product allocation decisions represent degree of process flexibility. We consider two scenarios: balanced and unbalanced demand and capacity. Analytic results and extensive computational results show that value of flexibility depends on demand and supply balance and number of facilities.

3 - Risk Sharing in a Coordinated Supply Chain under Considering

Quality Fault

Yongfei Li, School of Management, Xiían Jiaotong University,

No.28, Xianning West Road, Shaanxi, Xi’An, 710049, China, [email protected], Qin Su

Based on Stackelberg game, this paper presents the optimal profit, optimal order quantity and risk sharing problem in a coordinated supply chain under considering quality fault when the supply chain is consisted by one supplier and one retailer around a single product within one period. The risk sharing of all parties in the coordinated supply chain are all positively correlated with the quality fault rate, scrap rate, wholesale price and retail price but non-correlated with the return policy.

4 - A Decision-Theoretic Framework for Demand Estimation and

Inventory Planning during Hurricanes

Paul Cronin, University of Texas at Austin, 2110 Speedway Stop

B6500, CBA 5.202, Austin, TX, 78712-1750, United States of

America, [email protected], Douglas Morrice, John Butler

As a hurricane nears the Gulf coast, consumers stock up to prepare for the storm and its aftermath. Predicting consumer demand (and thus inventory allocation and planning) is a central challenge for retailers in the region. An econometric model estimates demand using actual purchase data and the residuals are used to explain consumer behavior using NOAA data from 2002-2008. A stochastic inventory allocation model investigates the timing and level of inventory allocation for future storms.

TB45

45- North 229 A- CC

Panel Discussion: Time Management

Sponsor: Junior Faculty Interest Group

Sponsored Session

Chair: Shengfan Zhang, Assistant Professor, University of Arkansas,

4207 Bell Engineering Center, Fayetteville, AR, 72701,

United States of America, [email protected]

1 - Panel Discussion: Effective Strategies for Time Management and Work-life Balance

Moderator: Shengfan Zhang, Assistant Professor, University of

Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, 72701,

United States of America, [email protected], Panelists:

Brian Denton, Mark Lewis, Laura McLay, Kathryn Stecke,

Ariela Sofer

In this discussion, panelists from engineering and business departments will share their effective strategies for time management and work-life balance. Panelists have experiences in winning NSF CAREER proposals, publishing in prestigious journals, mentoring a large research group, and serving at various committees while balancing and enjoying life.

300

TB46

46- North 229 B- CC

Disaster and Emergency Response II

Contributed Session

Chair: Soumia Ichoua, Associate Professor, Embry-Riddle Aeronautical

University, Department of Business Administration, Daytona Beach, FL,

United States of America, [email protected]

1 - Inventory and Routing Problem for Humanitarian Reliefs

Hoda Atef Yekta, Amirkabir University of Technology

(Tehran Polytechnic), Hafez St., Tehran, Iran, [email protected], Hamed Ahangari

This research proposed a mixed integer programming model to schedule the humanitarian relief vehicles after disaster. The objective of the model is finding the shortest routes for vehicles while considering the constraints of inventory at the main warehouse and small distributers besides considering the other conditions of disaster area.

2 - A Markov Decision Model for Capacity-adjustment in

Humanitarian Logistics Operations

Marco Serrato, Dr., Tecnologico de Monterrey, E. Monroy

Cardenas 2000. San Antonio Bue, Toluca, 50110, Mexico, [email protected], Jose Holguin-Veras, Raul Heras

The management of crisis situations through humanitarian aid operations has historically been identified as a critical issue. Unexpected events pose major challenges to the organizations involved in the delivery of critical supplies to an impacted site. A Markov decision model is developed through this research, which supports capacity-adjustment decisions for a point of distribution at the impacted site. The model allows the identification of an optimal monotone nondecreasing policy.

3 - Mitigation Strategies in Disaster and Crisis Management

Sushil Gupta, Professor, Florida International University, RB 250,

11200 SW 8th Street, Miami, FL, 33199, United States of America, [email protected], Martin Starr

We demonstrate the strategic importance of mitigation in disaster and crisis management where the responders must cope with needs of people buried, injured, starving, thirsty and requiring shelter. Crisis management anticipates needs before they arise. From a systems perspective, the anticipation of catastrophes can sometimes result in mitigation of damage. Under some circumstances, it can even prevent the catastrophe from occurring.

4 - A Stochastic Phase-dependent Approach for the Prepositioning

Problem in Humanitarian Supply Chains

Soumia Ichoua, Associate Professor, Embry-Riddle Aeronautical

University, Department of Business Administration, Daytona

Beach, FL, United States of America, [email protected],

Walid Klibi, Alain Martel

We propose a scenario based approach to tackle the strategic problem of prepositioning emergency supplies. First, the disaster process is modeled using a risk modeling approach and a Monte Carlo procedure is derived to generate a set of plausible disaster scenarios. Second, the problem is formulated as a two-stage stochastic programming model and solved using an SAA method based on the proposed Monte Carlo procedure. Experiments are conducted on a test case inspired from real-world data.

5 - Location of Volunteer Fire Departments in an Urban Region

Brigitte Werners, Prof. Dr., Ruhr-University Bochum, Institute of

Management, Fac of Management and Economics, Bochum,

44780, Germany, [email protected], Dirk Degel

Timeliness is the most important objective to reflect the quality of emergency services such as firefighting systems. In Germany, a well established system of volunteer fire brigades supports municipal fire departments. To optimize number and locations of volunteer fire departments multi-criteria methods are used to determine solutions for an urban area. Considered objectives are minimizing average travel time and longest travel time which both are in conflict with minimizing costs.

INFORMS Phoenix – 2012

TB48

TB47

47- North 230- CC

Dynamic Traffic Assignment II - System

Optimal Modeling

Sponsor: Transportation Science & Logistics/ Intelligent

Transportation Systems (ITS)

Sponsored Session

Chair: Tao Yao, Pennsylvania State University, 349 Leonhard Building,

University Park, PA, 16802, United States of America, [email protected]

1 - Generalized Point Queue Model (GPQM) and Application to

System Optimal Dynamic Traffic Assignment (SO-DTA)

Tao Yao, Pennsylvania State University, 349 Leonhard Building,

University Park, PA, 16802, United States of America, [email protected], Terry Friesz, Ke Han

The Vickrey’s point queue model involves an ODE with discontinuous right hand side, which makes it difficult to analyze and compute in continuous-time. We present a partial differential equation formulation of Vickrey’s model. This formulation leads to an explicit solution of the ODE and a new computational method. We will also discuss the application of the GPQM to SO-DTA problems.

2 - Modeling of Dynamic System Optimal in Continuous-Time

Rui Ma, Rensselaer Polytechnic Institute, Troy, NY,

United States of America, [email protected], Jeff Ban, Jong-Shi Pang

We propose an optimal control model for dynamic system optimal problem in continuous time. The model integrates the double queue formulation for queue dynamics that captures the queue capacity and inflow/exit flow capacities.

3 - On the Path Marginal Cost of System Optimal Dynamic Traffic

Assignment Problem

Hong Zheng, Research Assistant Professor, University of Arizona,

1209 E. Second Street, Tucson, AZ, 85721-0072, United States of

America, [email protected], Xianbiao Hu, Yi-Chang Chiu

In this talk we investigate and show a method to compute path marginal cost, on a general network with diversion points, for the system optimal dynamic traffic assignment (SO-DTA) problem. The method offers an effective way to solving the

SO-DTA problem. Some preliminary numerical results are presented.

4 - Dynamic System Optimal for Multiple O-D Networks Embedding

Cell Transmission Model

Kien Doan, Purdue University, 550 Stadium Mall Drive, West

Lafayette, IN, 47906, United States of America, [email protected],

Satish Ukkusuri

In this paper, the route-choice System Optimal Dynamic Traffic Assignment (SO-

DTA) problem for general network of multiple O-D pairs is formulated. We incorporate the Path-based Cell Transmission Model as the underlying network loading procedure to propagate traffic from different paths of different O-Ds. This model guarantees that the traffic moves without holding-back phenomenon.

Since the general optimization solvers are not able to solve the SO-DTA problem for medium-size network, we propose a heuristic algorithm based on method of successive average which can solve the SO-DTA problem for general networks.

Extensive experiment results will be shown to illustrate the benefits of the proposed model and algorithm.

TB48

48- North 231 A- CC

Facility Logistics I

Sponsor: Transportation Science & Logistics/ Facility Logistics

Sponsored Session

Chair: Ananth Krishnamurthy, University of Wisconsin-Madison, 1513

University Avenue, ME 3258, Madison, WI, United States of America, [email protected]

1 - Performance Analysis of Zone-picking Systems

Ivo Adan, Eindhoven University of Technology, Den Dolech 2,

Eindhoven, Netherlands, [email protected], Jelmer van der Gaast,

René De Koster, Jacques Resing

We consider zone picking systems, where totes travel between zones to collect items. If a tote tries to enter a fully occupied zone, then the tote is blocked and recirculated in the network. This system is approximated by a closed multi-class queueing network with jump-over blocking. We develop an iterative algorithm based on mean value analysis to evaluate blocking probabilities and performance characteristics such as utilization and throughput.

301

TB49

INFORMS Phoenix – 2012

2 - An Axiomatic and Teachable Warehouse Design Method

Leon McGinnis, Professor Emeritus, Georgia Institute of

Technology, Atlanta, GA, 30332, United States of America, [email protected]

Distinguishing function from embodiment is the key to developing a teachable warehouse design method that incorporates both analytic and empirical knowledge and is a platform for creating computer aided warehouse design tools.

The method will be described and illustrated with a case from the literature.

3 - Solving the Unequal Area Facility Layout Problem using Linear

Programming Based Genetic Algorithm

Sadan Kulturel-Konak, Associate Professor, Pennsylvania State

Berks, Tulpehocken Rd. P.O. Box 7009, Reading, PA,

United States of America, [email protected], Abdulah Konak

In this study, a hybrid Genetic Algorithm (GA)/Linear Programming (LP) approach is proposed to solve the FLP on the continuous plane with unequal-area departments. Once relative department positions are set, actual department locations and shapes are determined by solving an LP problem. Finally, the output of the LP solution is incorporated into the encoding scheme of the GA. Numerical results show that the proposed approach is very promising.

4 - Modeling Overlapping Container Terminal Operations

René De Koster, Erasmus University, Burgemeester Oudlaan 50,

Rotterdam, Netherlands, [email protected], Debjit Roy

We develop new integrated stochastic models for analyzing the performance of container terminals with overlapping loading and unloading operations. These models capture the complex interactions among the quayside, vehicle, and stackside processes, and provide system design insights.

TB49

49- North 231 B- CC

Planning and Routing in City Logistics

Sponsor: Transportation Science & Logistics

Sponsored Session

Chair: Teodor Crainic, Professor, ESG UQAM and CIRRELT, Pavillon

André-Aisenstadt, 2920, Chemin de la Tour, Montreal, QC, H3T 1J4,

Canada, [email protected]

1 - Efficient Routing on Large-scale Dynamic Networks under ITS using Hierarchical Communities

Mahyar Movahednejad, PhD Candidate, Wayne State University,

Manufacturing Engineering Building, 4815 Fourth St., Detroit, MI,

48202, United States of America, [email protected],

Ratna Babu Chinnam

We employ an emerging concept in analyzing complex networks called

“community structure detection” to capture traffic network dynamics in the form of hierarchical community-based representations of road networks. We then develop hierarchical routing algorithms to the extracted hierarchical network. We demonstrate the performance of the proposed approaches using actual road network data from Southeast Michigan.

2 - Urban Route Planning in the Management of Delivery

Time Windows

Jan Fabian Ehmke, University of Braunschweig,

Mühlenpfordtstrafle 23, Braunschweig, 38106, Germany, [email protected], Ann M. Campbell

Our goal is to improve the efficiency and reliability of last mile deliveries in urban areas. We present a computational framework, which considers time-dependent travel times in route planning of urban delivery tours. In particular, we investigate to which extent the design and the selection of delivery time windows might benefit from additional traffic information being available from telematicsbased data collection technology.

3 - Innovative Services in City Logistics

Luce Brotcorne, INRIA, 40 Avenue Halley-Bat A, Parc Scientifique de la Haute Borne, Villeneuve D’Ascq, 59650, France, [email protected], Alexandre Huart, Frederic Semet

Most providers in city logistics have transportation and storage capacities exceeding what is effectively used. To better meet the offer, an innovative strategy for an alliance of logistic providers is to develop a collaborative approach based on the pooling of residual capacities resulting of their activities. We propose a model for routing demands over a network defined by logistic services and show how innovative services can be modeled. Numerical results are presented.

4 - The Time-dependent Multi-zone Multi-trip Vehicle Routing

Problem with Time Windows

Phuong Khanh Nguyen, University of Montreal and CIRRELT,

2920 Chemin de la Tour, Montreal, QC, Canada, [email protected], Teodor Crainic, Michel Toulouse

The Time-dependent Multi-zone Multi-trip Vehicle Routing Problem with Time

Windows (TMZT-VRPTW), a problem arising from the second tier of the two-

302 tiered city logistic system described by (Crainic et al., 2009), is an extension of the

VRPTW involving both designing and assigning routes to vehicles within the time synchronization restrictions. We present a new approach which is able to improve the best known solutions of all currently published results.

TB50

50- North 231 C- CC

Modeling and Analyzing Military Operations and Systems I

Sponsor: Military Applications

Sponsored Session

Chair: Chris Arney, Professor, United States Military Academy,

Department of Mathematics, West Point, NY, 10996,

United States of America, [email protected]

1 - Network Cooperation Models for Irregular Warfare

Chris Arney, Professor, United States Military Academy,

Department of Mathematics, West Point, NY, 10996, United States of America, [email protected], Kathryn Coronges

We present a complexity-based modeling framework that merges network science and social network analysis with recent results in cooperative game theory and information science to simulate, design, and understand military systems and doctrine. This framework embraces the complexity of understanding the challenging issues associated with complex adaptive systems and military operations. We apply our framework to scenarios and important military operational issues.

2 - Packing Steiner Trees

Eric Thornburg, Department of Mathematical Sciences, United

States Military Academy, 646 Swift Road, West Point, NY, 10996,

United States of America, [email protected],

William Pulleyblank, Steven Horton

VLSI design requires efficient packing of Steiner trees linking components on a chip. We describe an algorithm for optimal generation and packing of Steiner trees in a rectilinear grid. “Good” Steiner trees are created by a cutting plane approach using Steiner multi-cuts. We use Benders’ decomposition to pack these

Steiner trees, while satisfying congestion constraints. We also discuss other applications of this approach.

3 - Social Media and its Future in the Army

Evan Szablowski, United States Military Academy, Department of

Mathematics, West Point, NY, 10996, United States of America, [email protected], Kathryn Coronges, Chris Arney,

Hilary Fletcher

Social Media (SM) have taken a central role in transforming the way people communicate and how we think about social engagement. Young generations have grown up enmeshed in SM, producing expectations about access to information. Young soldiers expect the military to be at the frontier of this revolution. We analyze how younger generations use SM and show how the military can benefit by applying the similar concepts to communication.

TB51

51- North 232 A- CC

Applying Analytics to Support Condition Based

Maintenance Decisions

Sponsor: Military Applications

Sponsored Session

Chair: Norm Reitter, Advisor, Information Technology, Concurrent

Technologies Corporation, 329, 44th Street, Pittsburgh, PA, 15201,

United States of America, [email protected]

1 - Using Sensor Data and Reliability Analysis to Improve Aircraft

Component Failure Predictions

Johnathon Dulin, Concurrent Technologies Corporation, 771

Fairdale Ct, Castle Rock, CO, 80104, United States of America, [email protected], Alan Moses

Sensor data use in Army Aviation is generally limited to determining when a component exceeds a maintenance threshold. With a wealth of data available from onboard sensors, those data could better support maintenance and sustainment decisions if used effectively. We use sensor data and operational profiles to project when a component will cross its thresholds, considering component and supply chain uncertainty to predict when a component will require replacement and to trigger a demand signal.

INFORMS Phoenix – 2012

2 - Using Large Scale Experiments and a CBM Simulation for

Driving Effectiveness in the Supply Chain

Dan Widdis, Fellow, Principal Operations Research Analyst,

Concurrent Technologies Corporation, 5897 Castleberry Peak Ave.,

Las Vegas, NV, 89131, United States of America, [email protected]

Component sensor data can be used both for maintenance decisions and to forecast future demand. Aggregated at the enterprise level, demand forecasts enable decision makers to prioritize orders, reposition stocks, and replenish stores in advance of anticipated future demands. Because of the large range of variables involved in these decisions and the potential interactive effects, we employ efficient experimental designs to extract the maximum information from a limited number of simulation runs.

3 - Using CBM Data for Army Aviation Fleet Sparing Optimization

Marshall Smith, Technical Director of Energy Systems, Clockwork

Solutions, 115 Wild Basin Rd S, Suite 301, Austin, TX, 78746,

United States of America, [email protected]

Maximum value of a CBM program in Army Aviation is realized when that program is tied to the Supply Chain supporting the fleet. We utilize the results of prognostic analytics at the component and tail number level to initialize all tail numbers within a fleet so that supply chain managers can optimize sparing strategies that best address the expected demands throughout the fleet, from the wholesale level down to the unit level.

TB52

52- North 232 B- CC

Modeling Natural Gas Markets with OR Techniques

Sponsor: Energy, Natural Res & the Environment/Energy

Sponsored Session

Chair: Olivier Massol, Assistant Professor, IFP School, 228-232 Avenue

Bonaparte, Rueil-Malmaison, 92852, France, [email protected]

1 - Natural Gas Price Forecasting via Selective Support

Vector Regression

Mohammad H. Poursaeidi, University of Houston, E 209

Engineering Building, Houston, TX, United States of America, [email protected], Erhun Kundakcioglu, Andrea Viacaba

Natural gas is one of the most abundant sources of energy in the US. An accurate natural gas price prediction model is useful in the industry due to its economical impacts. The purpose of this research is to analyze historical information on the variables that potentially have a high impact on the supply and demand for natural gas, as well as the price. This analysis is conducted using a novel datamining algorithm that is capable of simultaneously removing noise and performing regression.

2 - Joining the CCS Club! Insights from a Northwest European CO2

Pipeline Project

Stephane Tchung-Ming, Research Fellow, IFP Energies Nouvelles,

1-4 Avenue de Bois Preau, Rueil-Malmaison, 92852, France, [email protected], Olivier Massol

The large-scale diffusion of Carbon Capture and Storage (CCS) technologies imposes the construction of a sizeable pipeline infrastructure. We analyze the conditions for a widespread adoption of CCS by a group of emitters using a common pipeline system. It clarifies how the regulatory constraints imposed to the pipeline operator influence the emitters’ adoption. An application of this framework to the case of a European point-to-point pipeline is discussed. Policy recommendations are formulated.

3 - Cartelization in the Natural Gas Industry, a

Model-based Analysis

Olivier Massol, Assistant Professor, IFP School, 228-232 Avenue

Bonaparte, Rueil-Malmaison, 92852, France, [email protected], Albert Banal-Estañol, Steven Gabriel

We investigate the stability of a cartel gathering the main natural gas exporting countries. Our analysis uses a deterministic, discrete-time, finite-horizon oligopoly model capable to determine investment and production equilibrium strategies, in a setting where the demand’s specification takes into consideration the adjustment dynamics of current consumption to both present and past prices.

We assume a dynamic hierarchical pricing model á la Stackelberg with a feedback information structure.

4 - Analyzing Effect of U.S. Natural Gas Exports on the

Global Gas Markets

Steven Gabriel, University of Maryland, College Park, MD,

United States of America, [email protected], Seksun Moryade,

Hakob Avetisyan, Vincent Briat

U.S. natural gas companies aim to export natural gas to global markets because of growing domestic natural gas supply and relatively low prices compared to other foreign natural gas markets. In this presentation, we investigate effects of

TB55

exporting U.S. natural gas to Asia and Europe. We also take into consideration the effects of uncertainty of shale gas production in China in addition to carbon policy scenarios as part of University of Maryland’s World Gas Model.

TB53

53- North 232 C- CC

Stochastic Hydro-Thermal Scheduling II

Sponsor: Energy, Natural Res & the Environment/Energy

Sponsored Session

Chair: Steffen Rebennack, Assistant Professor of Operations Research,

Colorado School of Mines, 1500 Illinois St., Golden, CO, 80401,

United States of America, [email protected]

Co-Chair: Timo Lohmann, PhD Student, Colorado School of Mines,

816 15th Street, Golden, CO, 80401, United States of America, [email protected]

1 - Stochastic Dual Dynamic Programming with CVaR Risk

Constraints Applied to Hydrothermal Scheduling

Luiz Carlos Costa Jr, PSR, Praia de Botafogo 228/1701-A, Rio de

Janeiro, RJ, 22250-145, Brazil, [email protected], Mario

Veiga Pereira, Sèrgio Granville, Nora Campodonico, Marcia Fampa

In this talk we present an extension of the SDDP algorithm to incorporate CVaR constraints on supply reliability. We show that these constraints can be represented as piecewise linear penalties on the energy not supplied, and develop an efficient algorithm to calculate the segment coefficients. The methodology will be illustrated with a realistic stochastic hydrothermal scheduling problem.

2 - A Direct CVaR Approach with a Detailed Representation of

Critical Scenarios for the SDDP Method Applied to

Hydrothermal Generation Planning

Débora D. J. Penna, Researcher, CEPEL-Brazilian Electric Energy

Research Center, Avenida Horacio Macedo, 354, Ilha do Fundão-

Cidade Universitária, Rio de Janeiro, RJ, 22941-911, Brazil, [email protected], Maria Elvira Maceira, Andre Luiz Diniz

We present an alternative CVaR approach for the stochastic dual dynamic programming approach, with no additional state variables and a more intuitive recourse function, as compared to previous works. Aiming to improve the risk measure, we also perform a detailed representation of the tail of the distribution function by employing multivariate statistical techniques. Numerical results and discussion are presented in the multi-stage generation planning problem of the large-scale Brazilian system.

3 - Inflow Forecasting Models for Hydrothermal Scheduling

Timo Lohmann, PhD Student, Colorado School of Mines,

816 15th Street, Golden, CO, 80401, United States of America, [email protected], Amanda Hering, Steffen Rebennack

Inflow forecasting is an important part of hydrothermal scheduling under uncertainty. Algorithms such as the stochastic dual dynamic programming method (SDDP) require the forecasting model to be linear in order to maintain convexity in the future cost function. This limits the type of forecasting models and thus mostly simple methods such as linear autoregressive models are used in practice. We discuss these approaches and compare them to more advanced methods.

TB55

55- Regency Ballroom B - Hyatt

Undergraduate Operations Research Prize I

Cluster: Undergraduate Operations Research Prize

Invited Session

Chair: Feryal Erhun, Stanford University, Stanford, CA,

United States of America, [email protected]

1 - Detecting Covert Members of Terrorist Networks

Alice Paul, Harvey Mudd College, Claremont, CA,

United States of America, [email protected]

Increasing the amount of communication through a key leader increases the likelihood of detection. If we model a covert organization as a social network where edges represent communication between members, we want to determine the subset of members to remove that maximizes the amount of communication through the leader. We present an optimization and its decomposition for this problem. We focus on structural characteristics of vertices and subsets that increase communication.

303

TB56

INFORMS Phoenix – 2012

2 - Wind Turbine Offloading Strategy Optimization for GE Energy

Jacmara Ching, Georgia Institute of Technology, Atlanta, GA,

United States of America, [email protected], Santiago Diaz,

Antonio Elosua, Oscar Harasic, Yonatan Mintz,

Mario Solares Nassar

We created two tools to assist GE Energy with the wind turbine delivery and offloading processes. The first is a forecasting tool, which uses regression equations to forecast the arrivals of components at wind farms. The second is a sequencing tool, which utilizes a heuristic to find the optimal allocation of cranes at a given wind farm site. GE Energy can expect savings of $1.9 million from the implementation of these tools, and GE Energy’s customers may also expect savings up to this amount.

3 - Optimization-based Approaches to Assigning Sorority Sisters to

Rooms and Roommates

Xing Li, University of Michigan, Ann Arbor, MI, United States of

America, [email protected], Zixiao Chen, Rachel Froelich, Jon Lee,

Mark Daskin

Every semester, the Alpha Phi Sorority University of Michigan Chapter needs to assign enrolled sisters to different capacity apartments of the sorority in considering each individual’s room as well as roommate preference. We address this issue by using mixed-integer linear programming and heuristic algorithms to make optimal level assignment decisions that can be repeated every semester.

3 - The Test Region Iterative Contraction (TRIC) Method for Solving

Stochastic Control Problems

Chunyu Yang, Assistant Professor, BI Norwegian Business School,

Oslo, Norway, [email protected], Stathis Tompaidis

We introduce the test region iterative contraction (TRIC) method to solve highdimensional discrete-time stochastic control problems under general constraints.

The method combines dynamic programming and functional approximations of conditional expectations, and iteratively shrinks the approximation region around the optimal solution. We apply the method to two Finance applications: a) dynamic portfolio choice with constraints; b) dynamic portfolio choice with capital gain taxation.

4 - American Options: Upper Bounds on the Cost of

Delayed Exercise

Arun Chockalingam, Assistant Professor, Eindhoven University of

Technology, Den Dolech 2, Eindhoven, 5612AZ, Netherlands,

[email protected], Haolin Feng

In this talk, we consider the cost of delaying the exercise of American options.

Leveraging on the moving boundary approach to American option pricing, we characterize upper bounds on this cost in terms of only the option’s intrinsic value and the value function associated with delayed exercise for a variety of market models from the simple Black-Scholes model to complex Stochastic Volatility

Jump-diffusion models. We present illustrations of these bounds for both put and call options.

TB56

56- Curtis A- Hyatt

Joint Session TMS/NPD: Technology Management

Section Best Paper Winner Presentation

Sponsor: Technology Management & New Product Development

Sponsored Session

Chair: Cheryl Druehl, George Mason University, Fairfax, VA,

United States of America, [email protected]

1 - Technology Management Section Best Paper Winner

Presentation

Cheryl Druehl, George Mason University, Fairfax, VA,

United States of America, [email protected]

Please join us to award the first annual TMS Best Paper Award. This is for the best paper from 2007 in the TMS domain in an INFORMS journal. The Winner and

Runner-ups will present their work building on their seminal papers and specialguest discussants will comment.

TB57

57- Curtis B- Hyatt

Applications of Stochastic Control to Finance

Sponsor: Applied Probability

Sponsored Session

Chair: Stathis Tompaidis, University of Texas at Austin, 1 University

Station, B6500, Austin, TX, 78712, United States of America, [email protected]

1 - Algorithmic Trading: Matching the Volume Weighted

Average Price

Daniel Mitchell, University of Texas, McCombs School of Business,

1 University Station, B6000, Austin, TX, 78712, United States of

America, [email protected],

Jedrzej Bialkowski, Stathis Tompaidis

We consider a model in which a trader seeks to minimize the deviation between his volume weighted average price and the market’s, over a day for one stock.

The model incorporates both random fluctuations in volume and market impact of trading. We find some heuristic trading strategies as well as lower bounds on the minimal expected deviation. The model is calibrated to actual market data and we compare amongst the heuristic strategies using this real data.

2 - G-Brownian Motion and Optimal Stopping

Xin Guo, University of California-Berkeley, Berkeley, CA,

United States of America, [email protected]

We consider optimal stopping problems under the G-B.M. framework. Wellposedness of the problem and characterization of the solution are discussed.

Based on joint work with Chen Pan and Shige Peng.

TB58

58- Phoenix East- Hyatt

Data-driven Decisions: Methods and Applications

Sponsor: Applied Probability

Sponsored Session

Chair: Gah-Yi Vahn, Assistant Professor, London Business School,

Regent’s Park, London, United Kingdom, [email protected]

1 - Approximating the Crowd with Crowd Sense

Cynthia Rudin, Assistant Professor, Massachusetts Institute of

Technology, 77 Massachusetts Avenue, Cambridge, MA, United

States of America, [email protected], Seyda Ertekin, Haym Hirsh

We give an algorithm for “approximating the crowd,” which is to find the crowd’s majority vote on an item without paying all of them.

2 - Automated Risk Predictions for Clinical Decision Making

Mohsen Bayati, Stanford University, Stanford, CA, United States of

America, [email protected], Mark Braverman, Justin Gatewood,

Eric Horvitz

We present results on applications of automated machine learning methods in healthcare.

3 - Operational Statistics for the Risk-averse Newsvendor Problem

Mengshi Lu, University of California, 4141 Etcheverry Hall,

Berkeley, CA, 94720-1777, United States of America, [email protected], George Shanthikumar, Z. Max Shen

We consider the risk-averse newsvendor problem under the conditional value-atrisk criterion with unknown demand parameter. We use the operational statistics approach to integrate parameter estimation and optimization, and show that operational statistics achieves higher expected utility than the separated approach.

We also found that using operational statistics brings larger benefit when the risk associated with demand uncertainty and demand sampling is higher.

4 - Performance-based Regularization for Data-driven Optimization

Gah-Yi Vahn, Assistant Professor, London Business School,

Regent’s Park, London, United Kingdom, [email protected],

Noureddine El Karoui, Andrew Lim

We present an extension of the regularization technique that is currently popular in statistics and machine learning. We apply this to mean-CVaR portfolio optimization to reduce the estimation risk of the performance.

304

INFORMS Phoenix – 2012

TB59

59- Phoenix West- Hyatt

Queues with Abandonments

Sponsor: Applied Probability

Sponsored Session

Chair: John Hasenbein, Associate Professor, University of Texas at

Austin, 204 E. Dean Keeton St. Stop C2200, Austin, TX, 78712,

United States of America, [email protected]

1 - Comparing Ticket and Standard Queues in Heavy Traffic

Otis Jennings, Visiting Associate Professor, Columbia Business

School, 3022 Broadway, New York, NY, 10027, United States of

America, [email protected], Jamol Pender

A new customer to a ticket queue is offered a number, indicating order of service, and is told the number currently in service. The customer balks or joins. Even if he takes a ticket, he may abandon later, an event unbeknownst to the server until service would have begun. We use limit theory to compare ticket and standard queues, proving the two processes converge together to the same limit. We conclude that for balanced systems with patient customers, the approaches are relatively identical.

2 - Mean and Variance Control of Dynamic Rate Multiserver

Queues with Abandonment

Jamol Pender, Graduate Student, Princeton University,

Sherrerd Hall Princeton University, Princeton, NJ, 08540, United

States of America, [email protected], William Massey

We derive the optimal staffing schedule for a dynamic rate multi-server queue with abandonment where we want our profit to have a minimal amount of risk.

To this end we consider optimal staffing of our multiserver queue with the fluid mean, but also the diffusion variance. Our analysis yields that the optimal staffing level is the solution to a fixed point equation. We also compare our staffing schedule to more well known staffing procedures such as square root staffing.

3 - Dynamic Scheduling of a GI/GI/1+GI Queue with Many

Customer Classes

Jeunghyun Kim, University of Southern California, Marshall

School of Business, Los Angeles, CA, 90089, United States of

America, [email protected], Amy Ward

We consider a dynamic scheduling problem for a GI/GI/1+GI queue with many customer classes in which there is a class-dependent abandonment cost. The objective is to minimize average cost by dynamically choosing which class the server should next serve. We formulate and solve an approximating Brownian control problem with non-linear drift that incorporates the entire abandonment distribution of each customer class, and then interpret the solution as a scheduling control in the original problem.

4 - A Time-dependent Study of Markovian Queues with Balking and Reneging

Brian Fralix, Assistant Professor, Clemson University,

Department of Mathematical Sciences, Clemson, SC, 29634,

United States of America, [email protected]

We study the transient behavior of a birth-death process, whose birth and death rates are decreasing and increasing, respectively, with respect to the state index.

Such processes can be used to model Markovian systems where customers are allowed to either balk or renege. Our results are derived by relating this birthdeath process to a special type of queueing system, whose queue-length process is equal in distribution to the original birth-death process.

TB60

60- Remington- Hyatt

Airport Capacity Frontier Analysis: Current Methods,

Tools and Findings: Part II

Sponsor: Aviation Applications

Sponsored Session

Chair: Dipasis Bhadra, Senior Economist, USDOT/FAA,

800 Independence Avenue, Washington, DC, DC, 20009,

United States of America, [email protected]

1 - Airport Capacity using Non-parametric Methods; An Application to JFK Airport

Dipasis Bhadra, Senior Economist, USDOT/FAA, 800

Independence Avenue, Washington, DC, DC, 20009,

United States of America, [email protected], Steve Stroiney,

Michael Carpenter, Yu Zhang

As air traffic returns to its pre-recession levels, estimation of airport capacity for efficient planning of landing and takeoff and mitigation of congestion-induced delays, especially at nation’s busier airports, have become critical once again.

TB61

Approaching representative airport’s capacity, however, in parametric fashion

(i.e., analytic and numerical methods) provides an abstraction that, in many cases, may not be equivalent to tactical and the observed capacity. Using ASDE-X and combined with other publicly available data, we present an analytical and non-parametric empirical framework to estimate and project airport capacity that takes into account dependencies of runways to gate, taxiways and movement areas and airport systems; observed aircraft mix, types of queuing and other ATM requirements, weather conditions and runway characteristics.

2 - Airport Capacity Assessment using Advanced

Econometrics Models

Yu Zhang, University of South Florida, Civil and Environmental

Engineering Dept, 4202 East Fowler Ave., Tampa, FL, 33620,

United States of America, [email protected]

Airport capacity varies with many factors, to name a few, the configuration of runways, the fleet mix of the flights, weather condition. It also show different values when controllers trade off arrival and departure capacities to accommodate imbalanced demand in air traffic flow management. This study applies a combined Tobit regression model and Vector Autoregression Model to analyze the airport capacity and understand the incremental effects from NextGen programs deployed in recent years.

3 - Performance Evaluation of Alternative Continuous

Descent Approaches

Jens Brunner, TUM School of Management, Arcisstrasse 21, LS

Tech Dienstl Oper Mgmt, Munich, 80333, Germany,

[email protected], Elsayed Elsayed

Continuous Descent Approach (CDA) is an important segment of an aircraft under the new NextGen air traffic system. We present a new formulation for the problem in order to minimize fuel consumption, emissions, noise and delays subject to assignment and separation constraints. We evaluate different CDA approaches, i.e. non/some/all aircrafts are CDA equipped. Computational results using real data show the efficiency of the modeling and the potential benefits of the CDA implementation.

4 - Safety Analysis of Centralized and De-centralized Automated

Separation Concepts

John Shortle, Associate Professor, George Mason University,

Dept. of Systems Engineering & OR, Fairfax, VA, 22030,

United States of America, [email protected]

The Advanced Airspace Concept is a proposed concept involving centralized ground stations that automatically detect conflicts and provide resolutions to aircraft. In an alternate concept, governed by Autonomous Flight Rules, conflicts are detected and resolved in a de-centralized manner by individual aircraft. This talk explores high-level tradeoffs between such systems from a safety and reliability perspective, considering the failures of individual components and their potential impacts on safety.

TB61

61- Russell- Hyatt

Risk-Based Decision Making in Emergency Situations

Cluster: Applications in Emergency Management and

Terrorism Security

Invited Session

Chair: Igor Linkov, US Army Engineer Research and Development

Center, 696 Virginia Road, Concord, MA, 01742,

United States of America, [email protected]

1 - Valuation in Comparative LCA- Laundry Detergent Case Study

Valentina Prado, Graduate Research Assistant, Arizona State

University, School of Sustainable Engineering, P.O. Box 875306,

Tempe, AZ, 85287, United States of America, [email protected], Thomas Seager, Steven Tylock,

Melissa Bernardo

To better manage environmental tradeoffs, weights and uncertainty, Stochastic multi-attribute analysis can be used in the valuation stage of comparative Life

Cycle Assessment. Stochastic Multi-attribute analysis allows for uncertainty in the input parameters and explores a variety of weights simultaneously. Thus, it can help decision makers view the problem from multiple perspectives.

305

TB62

2 - Emergency Preparedness and Robustness to Uncertain

Population and Workforce Behaviors

James Lambert, Research Associate Professor, University of

Virginia, P.O. Box 400747, Charlottesville, VA, 22904,

United States of America, [email protected],

Jose Orlando Gomes, Elizabeth B. Connelly

Emergency managers can have mistaken assumptions of future population and workforce behaviors, in terms of their personal preparedness and their actions in emergencies. It is essential to understand the robustness of emergency plans to the variety of such behaviors. This talk will describe several case studies that model and evaluate the implications of population and workforce behaviors for risk management and emergency management in washington, DC, and Rio de

Janeiro, Brazil.

3 - Use of Cognitive Modeling and Decision Analysis for

Risk Management

Igor Linkov, US Army Engineer Research and Development

Center, 696 Virginia Road, Concord, MA, 01742,

United States of America, [email protected]

Risk management challenges facing the US DOD require integrated approaches for knowledge acquisition and decision making. This paper illustrates the use of cognitive modeling and decision-analytical approaches to: (i) flood risk perception and management in US Army Corps of ENgineers; (ii) integration of cognitive models and multi-criteria decision analysis to address course of action selection in emergency settings.

TB62

62- Borein A- Hyatt

Auctions and Private Information

Cluster: Auctions

Invited Session

Chair: Dirk Bergemann, Douglass and Marian Campbell Professor of

Economics, Yale University, 30, Hillhouse Avenue, New Haven, CT,

06520, United States of America, [email protected]

1 - Extremal Information Structures of the First Price Auction

Dirk Bergemann, Douglass and Marian Campbell Professor of

Economics, Yale University, 30, Hillhouse Avenue, New Haven, CT,

06520, United States of America, [email protected]

We investigate the effect of information on the equilibrium behavior in first-price auctions. We solve for the Bayes correlated equilibria (BCE) of the first-price sealed-bid auction with private values. The set of BCE correspond to the set of pairs of information structures and Bayesian equilibria. We use the duality to back out information structures that support equilibria on the frontier of bidders’ payoffs and revenue.

2 - Dynamic Revenue Maximization: A Continuous Time Approach

Philipp Strack, Department of Economics, University of Bonn,

Bonn, Germany, [email protected], Dirk Bergemann

We characterize the profit-maximizing mechanism for repeatedly selling a nondurable good in continuous time. The valuation of each agent is private information and changes over time. At the time of contracting every agent privately observes his initial type which influences the evolution of his valuation process. In the profit-maximizing mechanism the allocation is distorted in favor of agents with high initial types.

3 - A New Approach to Correlation of Types in Bayesian Games

Luciano De Castro, Northwestern University,

2001 Sheridon Road, Evanston, IL, United States of America, [email protected]

We propose a new approach to the model of correlation of types in Bayesian games, which also allows asymmetries. This is related to the idea that “beliefs do not determine preferences,” and consists of modeling types with two explicit parts: one for preferences and another for beliefs. Building on this idea, we are able to provide the first pure strategy equilibrium existence for a general model of multi-unit auc- tions where types can be correlated.

4 - Rational Market Making with Probabilistic Knowledge

Abe Othman, Carnegie Mellon University, Computer Science Dept,

Pittsburgh, PA, United States of America, Tuomas Sandholm

We study a market maker (MM) that has a prior over future world states and of how the price of a bet affects the probability that traders will accept it. Computing the optimal policy for a risk-neutral MM is simple, but challenging for a Kelly criterion MM. The former does not depend on the MM’s state, while the latter depends intricately on time and state. Counterintuitively, the latter may offer bets that are myopically irrational with respect to her beliefs for the entire trading period.

INFORMS Phoenix – 2012

TB63

63- Borein B- Hyatt

Behavioral Issues in Revenue Management

Sponsor: Behavioral Operations

Sponsored Session

Chair: Anton Ovchinnikov, Darden School of Business,

University of Virginia, 100 Darden Blvd, Charlottesville, VA, 22903,

United States of America, [email protected]

1 - Are Consumers Really Strategic? Experimental Study

Nikolay Osadchiy, Emory University, Atlanta, GA, United States of

America, [email protected], Elliot Bendoly

Novel retail selling mechanisms involve a tradeoff between buying now vs later at a lower price. Recreating this setting in a laboratory, we observe a substantial fraction of subjects making decisions consistent with expected utility maximization. We find that subjects are particularly sensitive to whether the information about riskiness of the buy-later option is provided. Implications for revenue management are discussed.

2 - Consumer Loyalty, Strategic Behavior and Retailers’ Dynamic

Pricing Strategy: Theory Sand Experiments

Benny Mantin, University of Waterloo, Waterloo, ON, Canada, [email protected], Mirko Kremer, Anton Ovchinnikov

We investigate the impact of strategic consumer behavior and loyalty on the dynamic pricing strategies of competing retailers. We present both modeling results as well as the results of behavioral experiments in which subjects played as retailers in either a monopoly or duopoly environments. We compare subjects’ decisions with the optimal model predictions and explain the effects of consumer loyalty and strategic behavior.

3 - Revenue Management with Lifetime Value Considerations

Anton Ovchinnikov, Darden School of Business, University of

Virginia, 100 Darden Blvd, Charlottesville, VA, 22903, United

States of America, [email protected], Phil Pfeifer

We discuss the interaction between revenue management (RM) and customer relationship management (CRM) for a firm that operates in a customer retention situation but faces limited capacity. We present a dynamic programming model for how the firm balances investments in customer acquisition and retention, characterize the optimal policy and discuss how it changes depending on capacity limitations. We then contrast the modeling results with those of a behavioral experiment.

4 - Revenue Management at a Theater: Estimating Time of Ticket

Purchases for Competing Customer Types

Necati Tereyagoglu, Assistant Professor, Georgia Institute of

Technology, College of Management, Atlanta, GA,

United States of America, [email protected],

Senthil Veeraraghavan, Peter Fader

Using individual level ticket sales and seat pricing data for a season of 21 concerts from a well-known arts organization, we test for the impact of discounts, show quality, performance date, promotions and time related characteristics on timing of the customer’s purchases. We identify attractive ticket price tiers for subscribers and occasional buyers over time; and develop discount decisions that increase the revenues from each performance.

TB66

66- Ellis West- Hyatt

Data Mining Applications

Sponsor: Data Mining

Sponsored Session

Chair: Rong Duan, Principle Member of Tech Staff, AT&T Labs,

180 Park Ave, Florham Park, NJ, 07932, United States of America, [email protected]

1 - Guess What I’m Doing Now!? An Analytic Engine For Detection of Mode-of-motion in Humans

Saeed Ghassemzadeh, Att Labs-Research, 180 Park Avenue,

Florham Park, NJ, 07932, United States of America, [email protected]

We introduce a generalized analytic engine which provides an algorithmic framework for detection of different modes-of-motion in human beings. The engine in turn can be used for detecting the deviation from the normal pattern and consequently correcting the anomalies. The algorithm was developed using the geriatric data from actual human subjects. The data was collected from

ZigBEE enabled sensors embedded in the insoles of subject’s slippers.

306

INFORMS Phoenix – 2012

2 - Multivariate Copulas Model in Spatiotemporal Rare

Event Detection

Rong Duan, Principle Member of Tech Staff, AT&T Labs,

180 Park Ave., Florham Park, NJ, 07932, United States of America, [email protected]

Characterizing mobility network traffic is one of the most challenging tasks in mobility network planning and designing, especially when special events may occur in some areas. The spatial neighbour cell towers are highly correlated to each other, especially for the rare events. In this paper, we propose multivaraite

Copulas model to detect the rare event clusters base on the spatial temporal data collected from mobility traffic data.

3 - Algebraic Graph Theoretic Approach for Data-driven

Characterization of Nano-scale Surface Features

Prahalad Rao, Oklahoma State University, 322 Engineering North,

Stillwater, OK, 74078, United States of America, [email protected], Satish Bukkapatnam,

Zhenyu (James) Kong, Adam Fields, Ranga Komanduri

We present a spectral graph theoretic approach to characterize the morphology of surfaces finished using chemical-mechanical planarization (CMP) processing to near-optical quality.The proposed method is based on transforming the surface topography data into a fully connected graph, and extracting spectral properties of the Laplacian of the graph as features. The resulting Eigen analysis correlates (p ~

80%) well with surface defect density.

TB67

67- Ellis East- Hyatt

Nanomanufacturing and Nanoinformatics II

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Qiang Huang, Associate Professor, University of Southern

California, 3715 McClintock Avenue, GER240, Los Angeles, CA, 90089,

United States of America, [email protected]

Co-Chair: Lijuan Xu, University of Southern California, 3715

McClintock Ave, GER 236, Los Angeles, CA, 90089,

United States of America, [email protected]

1 - Real Time Monitoring of Ultrasonic Cavitation in the

Casting Process

Jianguo Wu, University of Wisconsin, 1513 University Avenue,

ME3255, ISyE, Madison, 53706, United States of America, [email protected], Shiyu Zhou, Xiaochun Li

In the research of Metal-Matrix-Nano-Composite (MMNC), ultrasonic cavitation is usually used to disperse the added particles, degas the molten metal and refine the grains in the casting process. The objective of this research is to develop a new method for real time monitoring of ultrasonic cavitation and dispersion status in the casting process. By studying the cavitation noise, a new method is proposed to choose the cavitation power and control the cavitation time.

2 - Interaction-driven Physical Model Selection for SA-MOCVD

Synthesis Process of Nanowires

Lijuan Xu, University of Southern California, 3715 McClintock

Ave., GER 236, Los Angeles, CA, 90089, United States of America, [email protected], Qiang Huang

Numbers of physical models are proposed to describe the SA-MOCVD synthesis of nanowires without definite conclusion. In this paper, we incorporate interaction modeling to capture the local variability and develop an interaction-driven technique to select physical models to achieve better understanding of the process.

3 - A Stochastic Model for Estimating the Yield of ZnO Nanowire

Growth Process

Haitao Liao, Associate Professor, University of Arizona, Tucson, AZ,

85721, United States of America, [email protected],

Jingyan Dong, Yong Zhu

In this research, the growth process of ZnO nanowire arrays is studied. Based on the image data collected at different sites of a substrate, we develop a spatial compound Poisson model to estimate the yield of the growth process.

4 - Towards Coarse-Grained Potential Development in Atomistic

Simulation for Nanostructure Synthesis

Satish Bukkapatnam, Professor, Oklahoma State University,

322 Engineering North, Stillwater, OK, 74078, United States of

America, [email protected], Changqing Cheng

A coarse-grained atomistic simulation approach is essential for investigating nanostructure synthesis process after the initial nucleation stages due to the high computational overhead involved in classical atomistic simulations (e.g., Monte

Carlo). We developed the effective potentials for the coarse-grained models.

Simulations show that the coarse-grained approach can save over 50% of Monte

Carlo simulation time.

TB68

68- Suite 312- Hyatt

Asset Valuation and Risk Modeling

Sponsor: Financial Services Section

Sponsored Session

TB68

Chair: Jim Bander, National Manager, Decision Sciences / Risk

Management, Toyota Financial Services, 3200 W Ray Rd, Chandler, AZ,

85226, United States of America, [email protected]

1 - Pricing Asian Options via Compound Gamma and

Orthogonal Polynomials

Bacel Maddah, Associate Professor, American University of Beirut,

Bliss Street, Beirut, Lebanon, [email protected],

Hrayer Aprahamian, Joe Naoum-Sawaya

We develop two novel approximations (CG3 and CGn) for arithmetic Asian options. CG3 utilizes a compound Gamma distribution of the price average. It is calibrated by analytically matching the first three moments. CG3 outperforms many other approximations in the literature, on both accuracy and CPU time.

CGn expands CG3 utilizing the concept of orthogonal polynomials from Physics, and matches n > 3 moments. CGn produces mostly near-”exact” results within one second.

2 - Constructing Investor Risk Preferences from Data

Vishal Gupta, Massachusetts Institute of Technology,

77 Massachusetts Ave., E40-135, Operations Research Center,

Cambridge, MA, 02139, United States of America, [email protected], Dimitris Bertsimas

Specifying risk preferences is critical to financial applications; yet, preferences are unobservable. Thus, we ask investors to self-describe as “conservative” or “risky.”

In this work we take a novel, data-driven perspective. Using ideas from inverse optimization, we construct distortion risk measures that are consistent with an investor’s historical holdings. Computational evidence suggests the method is robust to data error and misspecification.

3 - Time-varying Beta and the Value Premium: Evidence from the

Varying-coefficient Single-index Model

Chaojiang Wu, PhD Student, Department of Operations and

Business Analytics, University of Cincinnati, 2925 Campus Green

Dr., Cincinnati, Oh, 45221, United States of America, [email protected], Hui Guo, Yan Yu

We investigate whether the conditional CAPM helps explain the value premium using varying-coefficient single-index model. Such modeling allows for nonlinear dependence of conditional beta on state variables and significant variables are selected exhaustively. We find conditional beta co-moves with unemployment and inflation, and the price-earnings ratio. The alpha is smaller for the conditional

CAPM than for the unconditional CAPM; nevertheless, neither model fully explains the value premium.

4 - Enterprise Risk Management and Capital Budgeting under

Dependent Risks: An Integrated Framework

Jing Ai, Assistant Professor, University of Hawaii, 2404 Maile Way

C305, Honolulu, HI, 96822, United States of America, [email protected], Tianyang Wang

Risk management and capital budgeting are two critical components of corporate decision making. They need to be considered jointly at the enterprise level because of their interactions through dependent risk exposures and other synergetic relationships. This paper develops an integrated framework for optimally coordinating the two functions toward the streamlined strategic goal of the enterprise in a multi-period setting. A hypothetical example is given for a financial services company.

307

TB69

TB69

69- Suite 314- Hyatt

Energy Risk

Cluster: Optimization in Finance

Invited Session

Chair: Denis Mazieres, Head of Methodology & Modelling,

LCH.Clearnet & Birkbeck, University of London, Aldgate House,

33 Aldgate High Street, London, EC3N 1EA, United Kingdom, [email protected]s.bbk.ac.uk

1 - Modelling and Forecasting Electricity Price Risk with

Quantile Regression

Sjur Westgaard, Associate Professor, Norwegian University of

Science and Technology, 7491 Trondheim, Norway, [email protected], Derek Bunn, Arne Andresen,

Dipeng Chen

We investigate the capability of quantile regression (QR) methods to model and forecast the tails of the spot price in the UK electricity market. QR allows us to explore how the underlying price drivers affect the different quantiles of the distribution. We demonstrate how lagged prices, prices of gas, coal and carbon, forecasts of demand and reserve margin in addition to price volatility influence the spot price distribution.

2 - Optimization of Real Options: Structural Estimation of

Switching Costs in Peak Power Plants

Stein-Erik Fleten, Professor, Norwegian University of Science and

Technology, Industrial Economics and Technology Mgm,

Alfred Getz v 3, Trondheim, NO-7491, Norway, [email protected], Knut-Harald Bakke, Jon Ragnar Viggen,

Erik Haugom, Carl Ullrich

We analyze the real options to shutdown, startup and abandon gas fired power plants. The plants’ status for a given year is reported to the US Energy

Information Administration as either operating, in standby or retired. We use a stochastic two-factor model for the spark spread process, and estimate the irreversible costs of switching by structural estimation. The proposed nonlinear optimization analysis also indicates the spark spread triggers and option values for the switching decisions.

3 - Role of Price Spreads and Reoptimization in the Real Option

Management of Commodity Storage Assets

Nicola Secomandi, Tepper School of Business, Carnegie Mellon

University, Pittsburgh, PA, 15213, United States of America, [email protected]

This paper shows that the real option management of commodity storage assets is easy when the asset is fast and frictionless. In contrast, analysis of the general case justifies the use of reoptimization based heuristics for computing near optimal policies and lower bound estimates on the storage value. Further, the fast and frictionless asset optimal value function yields closed form penalties based on price spreads that simplify the estimation of near tight dual upper bounds.

4 - A Kernel Based Approach to Storage and Swing Contracts

Valuations in High Dimensions

Denis Mazieres, Head of Methodology & Modelling, LCH.

Clearnet & Birkbeck, University of London, Aldgate House,

33 Aldgate High Street, London, EC3N 1EA, United Kingdom, [email protected]

This work expands upon the original work from Hubbert & Mazieres, “ Tensor of

Radial Basis Functions:...”, where the Tensor of Radial Basis Functions (TBF) was first introduced and from Boogert & Mazieres,” A Radial Basis Function Approach to Gas Storage Valuation”, 2011, where it was later applied in gas storage valuation in two dimensions. This work introduces multi-factor price and volume dimensions to storage and swing contract valuation by using Kernels: RBF and

TBF.

INFORMS Phoenix – 2012

TB70

70- Suite 316- Hyatt

Social TV and Video

Sponsor: Information Systems

Sponsored Session

Chair: Shawndra Hill, Assistant Professor, The Wharton School,

University of Pennsylvania, 3730 Walnut Street, Suite 500,

Philadelphia, PA, 19103, United States of America, [email protected]

1 - The Quest for Content: How User-generated Links Can

Facilitate Online Exploration

Gal Oestreicher-Singer, Tel Aviv University, Ramat Aviv, Tel Aviv,

Israel, [email protected], Shachar Reichman, Jacob Goldenberg

Online content are presented as product networks, where nodes are content pages linked by hyperlinks. Recently, websites have begun to offer social networks as well, creating a dual-network structure. We use data from YouTube and investigate the role of this dual-network structure in facilitating content exploration.We also present seven internet studies in which participants, browsing a YouTube-based website, are exposed to different conditions of recommendations.

2 - What’s the Buzz About: Analyzing Social Media Response to TV Ads

Shawndra Hill, Assistant Professor, The Wharton School,

University of Pennsylvania, 3730 Walnut Street, Suite 500,

Philadelphia, PA, 19103, United States of America, [email protected]

We study the relationship between Ad attributes and online Buzz. We link the attributes to buzz levels, sentiment and specific content.

3 - Social TV: Linking TV to Social Media and Sales

Adrian Benton, University of Pennsylvania, Philadelphia, PA,

United States of America, [email protected],

Shawndra Hill

“Social TV” is a term that broadly describes the online social interactions that occur between viewers while watching television. In order to better understand the relationship between television content, buzz on online social media, and ultimately viewership and sales, we examine the American television program,

The Voice. In this talk, we present analyses linking television content presented to buzz on Twitter and sales on ITunes.

TB71

71- Suite 318- Hyatt

Energy Information Systems

Sponsor: eBusiness

Sponsored Session

Chair: Wolf Ketter, Associate Professor of Information Systems,

Erasmus University, Rotterdam School of Management, Rotterdam,

Netherlands, [email protected]

1 - Designing Balancing Mechanisms for Energy Markets using

Realistic Customer Models

Konstantina Valogianni, PhD Candidate, Erasmus University,

Rotterdam School of Management, Rotterdam, Netherlands, [email protected], Wolf Ketter, Mathijs de Weerdt, John Collins

We present balancing mechanisms for a decentralized and deregulated energy market using controllable customer capacities. We model customers’ controllable capacities based on real-world data, since a crucial part for designing precise balancing mechanisms is a realistic customer representation.

2 - Autonomous Agent-based Decision-making for the Smart

Electricity Grid

Markus Peters, PhD Candidate, Erasmus University,

Rotterdam School of Management, Rotterdam, Netherlands, [email protected], Wolf Ketter, Maytal Saar-Tsechansky,

John Collins

The vision of a Smart Electricity Grid requires substantial advances in intelligent decentralized control mechanisms. We study a novel class of autonomous brokers that derive profit-maximizing strategies from interaction with their environment and demonstrate their performance in experiments with real-world data. Our work lays the foundation for innovative intermediation services that enable customer participation in the Smart Grid, and enhance the economic sustainability of power systems.

308

3 - Energy Informatics in Transportation Systems: Combining

Telematics Data with O.R.

Sudip Bhattacharjee, Associate Professor, University of

Connecticut, Storrs, CT, United States of America, [email protected], Alex Tung

Empty trailer trips (backhaul) lead to revenue loss, pollution, fuel consumption, and cost over a billion dollars each year. We use telematics data to match routes for “backhaul brokering”. Data analytics is used to create maps of trailer movement and frequent patterns. Subsequently, optimization provides tools to choose revenue enhancing and energy saving routes. This contributes to the emerging area of energy informatics where energy optimization and revenue are key to business sustainability.

4 - Auction Design for Local Reserve Energy Markets

Reinhard Madlener, Professor of Energy Economics and

Management, RWTH Aachen University, Mathieustrasse, Aachen,

52074, Germany, [email protected],

Christiane Rosen

We develop an auction mechanism that is designed for a local energy market. It is aimed at enabling regional trading of ancillary services, but can also be used for administering negotiation processes in virtual power plants or microgrids. In order to test the performance of the proposed auction mechanism, a simple multiagentbased simulation program has been devised. We find that the theoretical predictions hold in fact and competition quickly leads to price convergence.

TB72

72- Suite 322- Hyatt

Computational Stochastic Optimization in Energy II

Sponsor: Computational Stochastic Optimization

Sponsored Session

Chair: Ricardo Collado, Post-Doctoral Associate, Princeton University,

Department of Operations Res. & Financial Eng, Sherrerd Hall,

Charlton Street, Princeton, NJ, 08544, United States of America, [email protected]

1 - Approximate Dynamic Programming for an Energy

Storage Problem

Warren Scott, PhD, Princeton University, ORFE Departement,

Princeton, NJ, 08544, United States of America, [email protected], Warren Powell

We use approximate dynamic programming to solve an energy storage problem which combines wind turbines, an electricity storage device, and the electrical grid in order to satisfy a load. For approximate policy iteration we compare leastsquares Bellman error minimization, Bellman error minimization with instrumental variables, and projected Bellman error minimization. For direct policy search, we apply the approximate knowledge gradient framework to efficiently optimize the storage policy.

2 - Adaptive Convex Enveloping for Multidimensional Stochastic

Dynamic Programming

Sheng Yu, George Washington University, 1776 G Street NW,

Suite 101, Washington, DC, 20052, United States of America, [email protected], Enrique Campos-Nanez

Adaptive Convex Enveloping (ACE) is a powerful general purpose method for solving convex stochastic dynamic programs. With its optimization-oriented design, ACE easily handles large numbers of decision variables and constraints with the speed and reliability of convex optimization, and approximates the value function with error controlled everywhere. As a demonstration, we use ACE on electric vehicle battery station management and find the optimal policy for charging massive number of batteries.

3 - Dynamic Optimization of Threshold Risk Measures &

Applications to Energy Markets

Ricardo Collado, Post-Doctoral Associate, Princeton University,

Department of Operations Res. & Financial Eng, Sherrerd Hall,

Charlton Street, Princeton, NJ, 08544, United States of America, [email protected], Warren Powell

We introduce a new family of risk measures suitable for applications where the natural risk-averse model requires not to fall below a threshold. We call these

“threshold risk measures” and it arise naturally in problems related to energy markets and hedging. Our presentation discuss the threshold risk measures in the context of dynamic programming and an application in the trading of energy futures and spot contracts.

INFORMS Phoenix – 2012

TB73

TB73

73- Suite 324- Hyatt

Lot Sizing and Production Planning

Contributed Session

Chair: Ilke Bakir, PhD Student, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology,

765 Ferst Dr. NW, Atlanta, GA, 30332-0205, United States of America, [email protected]

1 - Primary Pharmaceutical Manufacturing Scheduling Problem

Subhash C. Sarin, Professor, Virginia Tech, Virginia Polytechnic and State Univ., 250 Durham Hall, Blacksburg, VA, 24061, United

States of America, [email protected], Hanif D. Sherali, Lingrui Liao

We address an integrated lot-sizing and scheduling problem in a pharmaceutical supply chain. Products from multiple levels are scheduled on parallel and capacitated bays for production in batches. Sequence-dependent setup times and costs are required between batches of different product families, inducing a traveling salesman type of substructure. We present a column generation-based method in concert with valid inequalities to address this problem.

2 - The Capacitated Lot Sizing Problem with Piecewise Concave

Production Costs

Esra Koca, Bilkent University, Department of Industrial

Engineering, Bilkent, Ankara, 06800, Turkey, [email protected], M. Selim Akturk, Hande Yaman

In this paper, we consider the capacitated lot sizing problem with piecewise concave costs and concave inventory holding costs. We develop a dynamic programming (DP) algorithm that solves this problem in polynomial time when the number of breakpoints of the cost function is fixed and the breakpoints are the same for all periods. To the best of our knowledge, this is the first algorithm that solves this problem in polynomial time. We test the DP with different MIP formulations.

3 - Production Planning and Scheduling with Sequence-dependent

Changeover Cost

Qingwei Li, Eastman Chemical, 200 S Wilcox, Kingsport, TN,

United States of America, [email protected], Dayana Cope

To minimize the total change-over cost in production, we developed a multi-line multi-product discrete lot-sizing model with sequence-dependent change-over cost.

4 - Production Planning Problem for Suppliers with Uncertain

Advance Demand Information

Nobuyuki Ueno, Professor, Prefectural University of Hiroshima,

1-1-71,Minami-ku,Hiroshima, Hiroshima, 734-8558, Japan, [email protected], Koji Okuhara, Takashi Hasuike

We model a supplier’s production planning problem with uncertain advance demand information called ‘Naiji’ in automobile industries. It is formulated as a stochastic programming problem which minimizes the cost subject to a probabilistic inequality under the condition that the inventory in 2 periods is correlated. We propose a new procedure to obtain an optimal solution on the

‘lattice’ without requiring the repeated evaluation of gradients of a probabilistic function containing integrals.

5 - Column Generation for Production Planning with

Alternative Routings

Ilke Bakir, PhD Student, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, 765

Ferst Dr. NW, Atlanta, GA, 30332-0205, United States of America, [email protected], Reha Uzsoy

We propose a column generation approach for production planning problems with a very large number of alternative routings. Computational experiments show that the proposed approach finds optimal solutions that use a very small number of the possible routings compared to the classical path-based approach.

309

INTERACTIVE SESSION

Tuesday, 12:30pm - 2:30pm

Interactive Session

West Ballroom Foyer, Level 300

Interactive Session

Cluster: Interactive Poster Session

Invited Session

Chair: Young-Jun Son, The University of Arizona, Systems and

Industrial Engineering, Phoenix, AZ, United States of America, [email protected]

Co-Chair: Burcu Keskin, University of Alabama, Huntsville, AL, United

States of America, [email protected]

Co-Chair: Jian Liu, University of Arizona, Tempe, AZ, United States of

America, [email protected]

1 - Establishing the Optimal Drain Tile Network Based on

Field Characteristics

Luke Holt, P.h.D. Student, North Dakota State University, 1201

Albrecht Boulevard, Fargo, 58102, United States of America,

[email protected], Nimish Dharmadhikari

Midwestern agricultural producers have recently been subject to high precipitation growing conditions. During this time, commodity prices have been great, and input costs have risen. These trends have sparked the installation of field drain tile at record rates. OR analysis is done to establish the best tile network.

2 - A Combined-Model Solution Approach for the HVRPTW

Ilknur Uludag, PhD Student, Department of Industrial and Systems

Engineering, Auburn University, Auburn, AL, 36849,

United States of America, [email protected]

Two new mathematical formulations for the Heterogeneous Vehicle Routing

Problem with Time Windows (HVRPTW) are proposed. One model features a combination of binary and continuous decision variables, while the other model relies on only binary decision variables with discretized time. While neither model is superior individually, a novel solution approach utilizing a combination of both models appears to be promising. Empirical results demonstrate the effectiveness of the proposed approach.

3 - No-Arbitrage Implied and Local Volatility Surfaces for Option

Pricing and Hedging

Jaewook Lee, Seoul National University, 599 Gwanak-ro,

Gwanak-gu, Seoul, Korea, Republic of, [email protected]

We propose a novel smoothing method to implement no-arbitrage constraints in estimating the implied and local volatility surfaces extracted from option prices using multiple local bandwidths. Through simulation on KOSPI200 index options, we find that implied and local volatility modeling under arbitrage-free conditions show better performance in terms of estimation, pricing, and hedging near the out-of-the-money and in the money options with short maturities.

4 - A Genetic Algorithm for Vehicle Routing Problem with

Simultaneous Pickup and Delivery

Fulya Altiparmak, Gazi University, Dept.of Industrial Engineering,

Ankara, Turkey, [email protected], Ismail Karaoglan,

Fatma Pinar Goksal

This study considers vehicle routing problem with simultaneous pickup and delivery (VRPSPD). Since the VRPSPD is an NP-hard problem, a hybrid solution approach based on genetic algorithm and variable neighborhood descent algorithm is proposed. The effectiveness of the proposed algorithm is investigated by an experiment on benchmark problem instances available in the literature.

The computational results show that the proposed algorithm competes with the heuristic approaches in the literature.

5 - Decision and Risk Analysis Based Interactive Project Evaluation

Approach on Oil Project

Omer Faruk Baykoc, Gazi University, Maltepe, Ankara, Turkey, [email protected], Mustafa Duygu

This study presents an integrated modeling approach for oil exploration and production projects. While constructing the model which reflects the real system best, all input and output factors have been included in the model with their specific distributions. Model outcomes derived as a result of study, which are used for project decisions, were tried to be included whole of uncertainties caused by each factor along the model.

INFORMS Phoenix – 2012

6 - Supplier Selection and an Application in the Baby Food Sector

Bahri Berker Ugur, Student/Gazi University, Asikveysel Mah.

322.Sokak 8/18 Mamak, Ankara, Turkey, [email protected],

Birsen Çevik, Mesut Kozan, Mürsel Dalyanoglu, Bahar Özyörük

In this study,for a baby food manufacturer, supplier selection problem is discussed and Axiomatic Design (AD) method is proposed for the solution of the problem.

Since, all of the factors that affect supplier selection process can not be denoted by numerical values, the evaluation of the firms has been done by using triangular fuzzy numbers and fuzzy AD has been used for the evalution.

7 - Improving Production and Design Processes through Advances in Experimental Design Analysis

Jonathan Linton, Professor, University of Ottawa, DMS 6108

Laurier Street East, Ottawa, ON, K1N 6N5, Canada, [email protected], Chris Gatti, Quanhong Jiang,

Mark Embrechts

Manufacturers rely on Taguchi Design of Experiments for improving processes. A study of 107 experiments assesses the use of PLS offering insights on: experimental quality, results and additional information that is obtainable. Data sets fell into four categories: appropriate sample, more data advisable, biased model and inappropriate model. In summary, PLS offers additional information that enhances the use of Taguchi DOE and guides further experimentation.

8 - Parallel Machine Scheduling Problem: An Application

Gul Didem Batur, Gazi University, Industrial Eng. Dept.,

Ankara, Turkey, [email protected], Serpil Erol

We have considered the parallel machine scheduling problem arising in the cable production department of a bus factory. In order to obtain the best part/machine assignments and balance the workloads, we have constructed a mathematical model and proposed a heuristic algorithm. Solutions of both methods are compared with respect to the problem objectives and the methods’ effectiveness.

9 - When Prospects Don’t Matter: Decision-Making under

Resource Constraints & Impact on Performance

Niharika Garud, PhD Candidate, IIM Bangalore, C/O FPM Office,

IIM Bangalore, Bannerghatta Road, Bangalore, KN, 560076, India, [email protected]

This paper presents an alternate view of decision-making in inherently uncertain and resource constrained conditions where individuals minimize their losses irrespective of prospects of outcomes. These choice patterns are inconsistent with both expected utility theory & prospect theory. We found that individuals developing hi-tech products *focus on loss minimization in terms of their assets and resources and *are consistently “risk averse” irrespective of variation in prospects of outcomes.

10 - Planning a Cross-trained Workforce in Retail Sector

Cèsar Augusto Henao Botero, Pontifcia Universidad Catòlica de

Chile, Vicuña Mackenna 4860, Macul, Santiago de Chile, Chile, [email protected], Juan Carlos Muñoz Abogabir

Retail workforce inflexibility is one of the sources for under and overstaffing. We propose to use cross-trained staff that may be assigned to different store’s departments. Our model determines how many employees are required by department, how many will be cross-trained, what skills should they have, and their time dependent assignment. Some policy conclusions are derived from data obtained from a large retail firm in South America.

11 - Pharmaceutical Supply Network Design: Risk-Averse or

Risk-Neutral?

Alireza Madadi, Clemson University, 110 Freeman Hall,

Department of IE, Clemson, SC, 29631, United States of America, [email protected], Julia Sharp, Kevin Taaffe,

Scott Mason, Mary E. Kurz

We investigate a supply network design with unreliable supply. We consider two types of decision making policies a risk-neutral decision-making policy and a riskaverse policy wherein, the decision-maker uses a CVaR approach to measure risk and to define worst-case scenarios. After introducing the underlying optimization models, we present computational analysis and statistical analysis to compare the results of the risk-averse and risk-neutral policies and provide several managerial insights.

12 - How Does Network Structure Influence Traceability?

Abigail Horn, PhD Student, Engineering Systems Division, MIT, 77

Massachusetts Avenue, Building E40-261, Cambridge, MA, 02139-

4307, United States of America, [email protected]edu, Xin Lu

We study the problem of finding the source of spreading through a network by modeling a contaminant in the food distribution network. We construct an estimator for the outbreak source and show how its ability to narrow down the source identification problem changes with the network’s structure. We propose a measure for traceability, or the overall ability to identify the source of spreading given any set of outbreak observations, based on entropy.

310

INFORMS Phoenix – 2012

13 - Scheduling Close Combat Air Support using Simulation

David Cornejo, North Carolina State University, 1207 Goldenview

Ct, Durham, NC, 27713, United States of America, [email protected]

Close Combat Air Support, support of ground missions using aircraft, is essential to modern military operations. Demand for such support is stochastic and thus difficult to plan appropriately. We present a simulation model that allows analysis of the effects of various schedules on aircraft utilization and service levels.

14 - An Integrated Discrete-Event/System Dynamics Simulation of

Breast Cancer Screening for US Women

Jeremy Tejada, Postdoctoral Student, University of Texas, 210

Calibre Chase Drive, Apartment 303, Raleigh, NC, 27609,

United States of America, [email protected]

The objective of this research is to develop, validate, and exploit a simulation modeling framework for evaluating the effectiveness of breast cancer screening policies in the near future (that is, over the period 2012–2020) for US women who are at least 65 years old. This work includes an examination of key components in the breast cancer screening process for older women, and an approach to defining and modeling those components using simulation.

15 - Assembling Balanced Incomplete Block Design Tests with

Mixed Integer Linear Programming

Pei-Hua Chen, National Chiao Tung University, 1001 Ta Hsieh Rd.,

HsinChu, 300, Taiwan-ROC, [email protected],

Daniel Junglas

The nationwide educational assessment in Taiwan is based on the structure of a balanced incomplete block design with more than one content area per block.

This study uses Mixed Integer Linear Programming to assemble sets of parallel booklets following a balanced incomplete block design.

16 - Method of Successive Modifications of Functional for

Transportation Problem with Warehousing

Vladimir Tsurkov, Prof., Computing Centre of Russian Academy of

Sciences, Vavilov str., 40, Moscow, 119333, Russia, [email protected],

Dmitry Kuzovlev, Alexander Tizik

Classical transportation problem is considered with additional warehouse points for suppliers and consumers. Main functional is augmented with expenses for transportation to these warehouses. Appropriate variables specifying amounts of goods delivered to warehouse points are introduced. Method of successive modifications of functional is employed for solving since standard method yields laborious procedures. Sample is presented, which illustrates efficiency of the algorithm.

17 - Antiviral Resistance and Mitigation of Pandemic Influenza

Sandro Paz, PhD Candidate/Associate Professor, University of

South Florida/Pontificia Universidad Catolica del Peru, 18125

Birdwater Dr., Tampa, FL, 33647, United States of America, [email protected], Alex Savachkin

We are building a large-scale simulation optimization framework for antiviral based mitigation of pandemic influenza. The model considers an oseltamivirsensitive strain and a resistant strain with low/high fitness cost. Mitigation strategies include treatment of symptomatic cases and chemoprophylaxis of preand post-exposure cases.

18 - The Volunteer’s Dilemma in Readmissions and Other Game

Theory Problems in Healthcare

Brendan Bettinger, Graduate Research Assistant, Northeastern

University, 360 Huntington Ave., Boston, MA, 02115,

United States of America, [email protected]

Healthcare coalitions and coordination across organizations are subject to the volunteer’s dilemma. Each participant benefits when maximum care is provided, but also would prefer others assume more of the responsibility. We develop probabilistic models to describe these dynamics and explore conditions affecting the failure to provide care provision. We also explore other applications of game theory in a healthcare setting, including the tragedy of the commons and fair division.

19 - Reliability Analysis of Systems Subject to Degradation and Shocks

Sara Ghorbani, PhD Student, Rutgers, Rutgers 23629 BP.O. WaY,

Piscataway, NJ, 08854, United States of America, [email protected], Elsayed A. Elsayed, Hoang Pham

Abstract System’s failure can be described by underlying degradation and random shocks. In this paper, we consider systems subject to both degradation and shocks and that the impact of their impact is dependent on the magnitude of shocks as well as the “health state” of the system. We develop expression for the total damage at any time and calculate the system reliability for a given threshold.

INTERACTIVE SESSION

20 - A Simulation Optimization Approach to Epidemic Forecasting

Elaine Nsoesie, Virginia Tech, 1880 Pratt Drive, RBXV,

Blacksburg, VA, 24060, United States of America, [email protected], Sara Shashaani, Richard Beckman,

Kalyani Nagaraj, Madhav Marathe

We introduce a simulation optimization (SIMOP) approach for real-time forecasting of the epidemic curve. The forecasting SIMOP procedure combines an individual-based model and the Nelder-Mead simplex optimization method.

Epidemics simulated over synthetic social networks representing Montgomery

County in Virginia, Miami, and Seattle, as well as the H1N1 pandemic in Los

Angeles estimated from incidence data in September 2009, are forecasted using this method.

21 - Detecting Changes in the Mean from Censored Lifetime Data using Profile Monitoring techniques

Ehsan Jafari-Shirazi, West Virginia University, 2900 University

Ave. Apt 4, Morgantown, WV, 26505, United States of America, [email protected]

Monitoring censored lifetime data in which life has a univariate distribution has been investigated in the literature. Sometimes, life is dependent on some explanatory variables where it can be characterized as a function. In such situations, it is desired to monitor the function or the profile used for modeling the relationship between the explanatory variables and the response (life). In this study, two methods for monitoring censored lifetime profiles are proposed.

22 - Real-Time Integrated Airport Surface Movement Controlling

Qing Wang, University of South Florida, 14215 Les Palms Circle,

Apt. 201, Tampa, FL, 33613, United States of America, [email protected], Yu Zhang

In this study, we propose a real-time integrated airport surface movement controlling model. Given a set of arrival and departure flights in a specified planning horizon, the model determines 4D trajectories of aircraft that minimize total taxi time and maximize runway throughput subject to the operational constraints and safety requirements.

23 - Properties of a General Family of Agreement Metrics

Stephen France, Assistant Professor, University of Wisconsin-

Milwaukee, P. O. Box 742, Milwaukee, WI, 53207,

United States of America, [email protected]

We describe a family of metrics for measuring agreement between sets of solution configurations. We describe extensions for measuring confidence intervals and investigating statistical hypotheses. We describe a set of software tools developed in MATLAB and R. We show how the agreement metrics can be used to evaluate dimensionality reduction methods, tune method parameters, and evaluate how solution configurations change over time.

24 - Multi-period Defensive Resource Considering Equity and

Possibly Non-strategic Attackers

Xiaojun Shan, PhD Candidate, University at Buffalo, The State

University of New York, 435 Bell Hall, Buffalo, NY, 14260,

United States of America, [email protected], Jun Zhuang

We study defensive allocation in a multi-period multi-target defender –attacker game. We also study cost of equity (fairness) in defensive resource allocation. We develop a novel model in which a government allocates defensive resources among multiple targets, while reserving a portion of defensive resources

(represented by the equity coefficient) for equal distribution. Moreover, we study the robustness of defensive resource allocations against a possibly non-strategic attacker.

25 - Detecting the Undetectable: When the Distinction between

2+2=4 and 2*2=4 Means Everything

Raisa slepoy, Statistician, Private, 6006 Madawaska Road,

Bethesda, MD, 20816, United States of America, [email protected]

Hand recorded data captured in the field of operations have increasingly become important for agencies relying on fast turnaround requests for information to make fund allocations to insure resources meet the mission objectives.These

decisions drive the success or failure of the outcome. We propose a method of correcting quantity data by extending Luhns algorithm to validate identification numbers, such as credit card and ISBN numbers.Using this method we identified a common type of error.

26 - Bayesian Optimal Learning for Budget Allocation in Crowdsourcing

Qihang Lin, PhD Students, Tepper School of Business, Carnegie

Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213,

United States of America, [email protected], Xi Chen,

Dengyong Zhou

We study the budget allocation in crowd labeling for classification where each instance can be labeled multiple times. Given a fixed amount of budget, a challenging problem is how to allocate the budget across instances with different intrinsic ambiguity so that the classification accuracy is maximized. We adopt the

Bayesian optimal learning framework and formulate the problem into a finitehorizon MDP. We further propose several approximate DP methods to efficiently solve the MDP.

311

INTERACTIVE SESSION

27 - Analysis of Buyback Operations for Commercial Vehicles

Zeynep Kirkizoglu, Sales Support Manager, MAN Truck & Bus

Turkey, Esenboga Havalimani Yolu 22.Km, Akyurt, Ankara, 06750,

Turkey, [email protected]

We analyze the potential buyback operations for the Turkish sales office of a commercial vehicle manufacturer. Buyback operations are currently not fully implemented, but are expected to be one of the major business lines in the future.

We especially analyze the combined pricing of buybacks, including extended warranty and maintenance contracts; and the related benefits and costs. This is in line with the overall approach of “providing transport solutions,” rather than simply selling vehicles.

28 - Facilitating Interoperability between Community Networks using Semantic Technologies

Vijayan Sugumaran, Professor, Sogang University, Department of

Global Service Management, Seoul, Korea, Republic of, [email protected]

This research focuses on developing a service-oriented architecture consisting of middleware services to allow community-based data and knowledge creation, organization, sharing and inter-operation. We have developed a methodology and a set of tools that utilize loosely coupled web services that promote interoperability within and across communities connected through an interoperability network. The contribution is the delineation of the processes

(work flows) and the needed infrastructure.

29 - Semidefinite Approximations for Economic Order

Quantity Problems

Carolina Saldana Cortes, Professor, Universidad Externado de

Colombia, Calle 12 N 1-17 Este, BOGOTA, Colombia, [email protected]

This work is a research contribution in the direction to finding new ways to solve

(S,s) type basic problems, as is the Economic Order Quantity (EOQ) one in an effective way. To that end alternative methods of semidefinite relaxation of dynamical programs using binary variables are proposed. In order to offer a quantitative approach as an illustrative case, the models developed are applied to data coming from a representative firm of the Colombian ceramics sector.

30 - Investigation of Hiring & Wage Decisions in the Workforce

Supply Problem

Oanh (Olivia) Hoang, University of Alabama, [email protected], John Mittenthal, Nathan Palmer,

Charles Sox

Minimizing labor cost while attracting enough workers is critical for a multifacility enterprise. The decisions are how many workers to hire, and how much to pay each one. This study investigates coordinated versus uncoordinated workforce supply decisions. The results indicate that coordinated decisions yield lower total labor costs.

31 - Improving School Choice: Keeping Communities Together Via

Correlated Lottery

Peng Shi, MIT Operations Research Center, 77 Massachusetts Ave.,

E40-149, Cambridge, MA, United States of America, [email protected]

In school choice, children submit a ranked list of schools they prefer, and schools submit priorities over children; a centralized algorithm makes the assignment.

Current school choice algorithms tend to disperse communities so kids do not go to school with others from their neighborhood. We improve community cohesion in school choice mechanism by implementing a correlated lottery, which is NPhard in general but we have a heuristic that does well in practice.

32 - Scatter Search for the Single Row Facility Layout

Ravi Kothari, Indian Institute of Management Ahmedabad, c/o

SAO Office, Dorm-21, Room-16,, Ahmedabad, 380015, India, [email protected], Diptesh Ghosh

The single row facility layout problem is the problem of arranging facilities with given lengths on a line, with the objective of minimizing the weighted sum of the distances between all pairs of facilities. The problem is NP-hard and research has focused on heuristics to solve the problem. We present four scatter search algorithms to solve large sized instances. Our computational experiments show that the algorithms generate better solutions to 15 of the 58 benchmark instances.

33 - Planning Pilot Training in Turkish Naval Force

Bülent «atay, Sabanci University, Fac. of Engineering & Natural

Science, Istanbul, 34956, Turkey, [email protected],

Tevfik Altinalev

We address the pilot training planning problem at Turkish Naval Air Force. Our aim is to determine weekly flight training plans such that the number of combatready pilots is maximized at the end of the planning horizon. We formulate a 0-1 mixed-integer linear program and discuss heuristic methods based on LP relaxation, Lagrangean relaxation, and Lagrangean decomposition. Our experimental results show that the Lagrangean decomposition with subgradient optimization provides good bounds.

INFORMS Phoenix – 2012

34 - using Simulation to Improve Yarn Manufacturing Throughput

Wei Zhao, Clemson University, 854 Issaqueena Trail, Apt. #202,

Hunter’s Glen, Central, SC, 29630, [email protected], Scott Mason,

Kevin Taaffe

Glen Raven Custom Fabrics identified the need to improve their yarn manufacturing throughput. While theoretical capacity of subprocesses could be calculated, determining the practical operating capacity was challenging, given the multiple subprocesses as well as the size and mix of orders. Using simulation to experiment with staffing assignments, machine and process configurations, the company estimates their cost savings, after implementing the process changes, to be US $2.4 million per year.

35 - Railway Freight Distribution Network Design and New

Applications in Turkey

Bahar Özyörük, Ass. Prof. Dr., Gazi University, Gazi University

Faculty Engineering, Dep. of Industrial Eng., Ankara, 06570,

Turkey, [email protected]

Railways to carry too much load is very important for the country’s economy. In this study, the current railways transportation in Turkey are evaluated. Location of to increase the share of freight transport.Emphasized the importance of the establishment of villages. Assessed alternative areas of logistics village.A model was established and model is solved by using LINDO Software Solver and the results are interpreted.

36 - Minimizing Emissions in Facility Location

Fan Jia, University of Minnesota, 111 Church Street S.E.,

Minneapolis, MN, 55455, United States of America, [email protected], John Gunnar Carlsson

We consider a continuous facility location problem in which our objective is to minimize the total emissions. We first analyze the limiting behavior of this model and give an asymptotically optimal configuration - Archimedean spiral. We show that the total emissions can be reduced by over 27% this way compared to the popular honeycomb configuration. Finally, we give a fast constant-factor approximation algorithm for finding the emissions-optimal placement of facilities in any convex polygon.

37 - Structure of Demand and Consistent Conjectural Variations

Equilibrium in a Mixed Oligopoly Model

Vyacheslav Kalashnikov, Assistant Professor, ITESM, Campus

Monterrey, Ave. Eugenio Garza Sada 2501 Sur, Monterrey, NL,

64849, Mexico, [email protected], Vladimir Bulavsky,

Mario A. Ovando Montantes, Nataliya Kalashnykova

A model of mixed oligopoly with conjectured variations equilibrium with a noncontinuous demand function is studied. The agents’ conjectures concern the price variations depending upon changes in their production output. Existence and uniqueness results for the conjectured variations equilibrium for any set of feasible conjectures are established. A notion of the interior equilibrium (based on a consistency rule) is introduced, and the existence theorem for the interior equilibrium is proved.

38 - Modeling the Evolution of Dependency between Demands, with Application to Inventory Planning

Amirhosein Norouzi, PhD Student, NC State University,

6900 Crescent Moon Ct., Apt 306, Raleigh, NC, 27606,

United States of America, [email protected], Reha Uzsoy

We show that the progressive realization of uncertain demands through successive forecast updates results in the evolution of the conditional demandcovariance in addition to the conditional demand-mean. A dynamic inventory model with forecast updates is used to illustrate an application of our method. We show that how the optimal inventory policy depends on conditional covariances.

39 - U. S. Department of Homeland Security’s Simulations of

Pedestrian Movement for Site Security

Doug Samuelson, President/Chief Scientist, InfoLogix, Inc., 8711

Chippendale Court, Annandale, VA, 22003, [email protected]

The U. S. Secret Service (USSS) has substantial modeling and simulation capabilities for event planning and personnel training. Recent projects are a stadium / arena Evacuation Planning Tool (EPT) and a Site Security Planning Tool

(SSPT), a smaller-scale, video game based simulator for planning and training.

Next steps include more realistic crowd behaviors, responses to toxic plumes or secondary attacks, and pursuing a single integrated all-purpose, all-scales model and interface.

40 - Integrating Supply Chain Network Model for Auto Industry in

Midwest using GIS

Yasaman Kazemi, North Dakota State University, P.O. Box 6050,

NDSU UGPTI Dept. 2880, Fargo, ND, 58108, United States of

America, [email protected], Eunsu Lee

This study develops a model for domestic automobile supply chain in the Midwest area based on minimum cost (distance) and integrates it into highway transportation network via GIS spatial interaction to predict the minimum flow from origins to destinations by choosing an optimal sequence of routes in order to optimize the total cost of supply chain network.

312

INFORMS Phoenix – 2012

41 - Experiences with Data Mining in Support of Radiation Portal

Monitoring at Border Crossings

Tom Burr, Los Alamos National Lab, Mail Stop F600,

Los Alamos, NM, 87545, United States of America, [email protected],

Michael Hamada

Radiation Portal Monitors screen vehicles to protect against illicit special nuclear material (SNM) such as Uranium, which emit neutrons and/or gammas. One complication is naturally occurring radioactive materials (NORM) which leads to nuisance alarms. This paper describes data mining applied to RPMs as related to background suppression; efficient methods to examine multiple candidate alarm options; smoothing spectral data; sensor drift, and pattern recognition to recognize common NORMs.

42 - Machine Learning and Algebraic Geometry for Network Data

Ian Dinwoodie, Assistant Professor, Portland State University,

Dept. Math and Stat, Portland, OR, 97201,

United States of America, [email protected]

Understanding interactions and steady states for biological networks from data of multivariate time series requires a range of statistical and mathematical tools.

Discretization, machine learning, and algebraic geometry together can be used for practical questions of node dependence and exclusion from basins of attraction.

43 - Tractable Approximations to Multi-Period Hydroelectric Power

Station Management

Limeng Pan, University of Illinois at Urbana-Champaign, 117

Transportation Building, 104 S Mathews Ave., Urbana, IL, 61801,

United States of America, [email protected]

We study a complex Hydroelectric Power Station operation problem. A series of tractable approximations using Linear Decision Rules in the perspective of robust optimization are developed to achieve both the tractability and reliability. In the numerical study, the world’s largest multi-purpose reservoir, the Three Gorges

Reservoir (TGR), is used to test the performance of our models. The optimal values from our tractable approximations are matching the benchmark optimal value from SDP.

44 - A Recycling Network Design for Multiple Products and

Multiple Material Types

Ertan Güner, Prof.Dr., Gazi University, Industrial Engineering

Department, 06570- Maltepe, ANKARA, Turkey, [email protected], Kemal Kalaykiran

Recycling is one of the most common recovery methods both in the application.

In this study, a recycling network is designed for multiple products and multiple material types. The components of the system taken into consideration are customer zones, potential collection facilities and recycling facilities. The problem is defined in detail and a mixed integer mathematical model is presented.

45 - Supply Chain Design Simulation

Sima Maleki, University of Tennessee, 1172 Keowee Avenue,

Knoxville, TN, 37919, United States of America, [email protected]

ARENA is used to design alternative supply chain networks and compare them with existing network to help a manufacturing company decide on their future model. Transportation cost, throughput, lead time, capacity and service level are key performance factors. Considering uncertainty in demand, a centralized production hub with local distribution hubs offers better service. Hierarchical decentralized production facilities which perform local distribution offers lower transportation cost.

46 - Multiscale Recurrence Analysis of Long-Term Nonlinear and

Nonstationary Time Series

Yun Chen, Mr., University of South Florida, 4202 E. Fowler

Avenue, ENB 118, Tampa, FL, 33620, United States of America, [email protected], Hui Yang

Recurrence analysis is an effective tool to characterize and quantify the dynamics of complex systems. However, recurrence computation is highly expensive as the size of time series increases. Few, if any, previous approaches have been capable of quantifying the recurrence properties from a long-term time series. This paper presents a novel multiscale framework to explore recurrence dynamics in complex systems and resolve computational issues for a large-scale dataset.

47 - Transmission Expansion Planning using an Linearized AC

Model

Hui Zhang, Research Associate, Arizona State University, 5th floor,

Room 519, 551 E. Tyler Mall, Tempe, AZ, 85281, United States of

America, [email protected], Hui Zhang, Gerald Heydt,

Vijay Vittal, Jaime Quintero

The motivation of this work is to develop a less relaxed network model based on which a better TEP result can be obtained. First, a novel approach to linearize the full AC network model is presented. Based on the linearized network model, a mixed-integer second order cone programing (MISOCP)-based TEP model is then proposed. The proposed TEP model includes the linearized reactive power, offnominal voltage magnitudes and network losses.

Tuesday, 1:30pm - 3:00pm

TC01

TC01

01- West 101- CC

Global Optimization in Graphs/Networks

Sponsor: Optimization/Global Optimization

Sponsored Session

Chair: Sergiy Butenko, Associate Professor, Texas A&M

University,Industrial & Systems Engineering, 3131 TAMU, College

Station, TX, 77843, United States of America, [email protected]

1 - A Branch-and-bound Approach for Maximum Quasi-cliques

Zhuqi Miao, PhD Student, Oklahoma State University,

808 N. Monroe, Apt. 32, Stillwater, OK, United States of America,

[email protected], Foad Mahdavi Pajouh,

Baski Balasundaram

Detecting quasi-cliques is an important tool for detecting dense clusters in graphs.

Quasi-clique detection has been accomplished using heuristic approaches in various applications of graph-based data mining. However, exact approaches for the problem are limited to two mixed integer programming formulations that were proposed recently. This article presents a new combinatorial branch-andbound algorithm for the maximum quasi-clique problem and studies its performance on benchmark instances.

2 - On a Class of Bilevel Linear Assignment Problems

Vladimir Stozhkov, University of Florida, 303 Weil Hall,

Gainesville, FL, 32611-6595, United States of America, [email protected], Vladimir Boginski, Eduardo Pasiliao,

Oleg Prokopyev

We consider a bilevel extension of the linear assignment problem. The follower solves an assignment problem maximizing his profit, while the leader may interdict removing some of the nodes of the graph. The leader’s task is to minimize the total cost given by the cost of interdiction plus the cost of the assignments made by the follower. The problem is shown to be NP-hard. We also derive equivalent linear mixed 0-1 programming reformulations and discuss possible heuristic solution approaches.

3 - Minimum Selective Dominating Set Problem

Je Sang Sung, Research Assistant, Texas A&M University,

302 Ball Street, Apt. G204, College Station, TX, 77840, United

States of America, [email protected], Sergiy Butenko

We introduce several variations of the classical graph-theoretic concept of domination that are motivated by practical considerations. Complexity of the corresponding optimization problems is analyzed showing these variations are hard on their own respect. We also establish some basic properties of the corresponding polyhedra, and analytical bounds on the size of structures of interest.

4 - A Global Optimization Algorithm for the Maximum Biclique

Problem in Large-scale Networks

Shahram Shahinpour, Texas A&M University,Industrial & Systems

Engineering, TAMU-3131, College Station, TX, 77843, United

States of America, [email protected], Zeynep Ertem,

Sergiy Butenko

We present a scale reduction technique for solving the maximum vertex biclique problem in general graphs. Analytical bounds for the solution size in random graphs will also be explored. Experimental results on large real life networks show that the proposed algorithm is very effective in solving all test cases to optimality.

313

TC02

INFORMS Phoenix – 2012

TC02

02- West 102 A- CC

Joint Session DAS/HAS: Decision Analysis in

Public Health

Sponsor: Decision Analysis & Health Applications Society

Sponsored Session

Chair: Jagpreet Chhatwal, Assistant Professor, University of Pittsburgh,

130 De Soto St, Pittsburgh, PA, United States of America, [email protected]

1 - Optimal Colorectal Cancer Screening: Balancing Harms and Benefits

Fatih Safa Erenay, Assistant Professor, University of Waterloo,

200 University Ave. CPH 4323, Waterloo, ON, Canada, [email protected], Oguzhan Alagoz, Adnan Said

We use a POMDP model to determine the optimal colonoscopy screening policies that balance the harms (increased costs, disutility, complication risk) and benefits

(increased QALYs, colorectal cancer risk) of screening using real data. We find that the optimal policies recommend more aggressive screening and some of the proposed policies outperform current guidelines in terms of QALYs and total costs.

We also compare the performances of colonoscopy and other screening methods using our model.

2 - Modeling Latent Therapeutic Demand for Common

Variable Immunodeficiency

Jeffrey Stonebraker, Assistant Professor, North Carolina State

University, Poole College of Management, Raleigh, NC,

United States of America, [email protected]

This research applies methods from decision analysis to model latent therapeutic demand for common variable immunodeficiency (CVID) by integrating the variability in epidemiological data and treatment modalities. This approach is novel to forecast potential demand for the treatment of CVID where supply limitations have reduced the usefulness of sales-based forecasting methods. This approach can help manufacturers and healthcare agencies to ensure adequate treatment for patients with CVID.

3 - Cost-effectiveness of Expanded HIV Treatment among Pregnant

Women in Ghana

Adam VanDeusen, Master of Public Health Candidate,

Yale School of Public Health, 234 Park Street, New Haven, CT,

06511, United States of America, [email protected],

Elisa Long, Elijah Paintsil, Thomas Agyarko-Poku

Current World Health Organization guidelines recommend short-term antiretroviral treatment for healthy, HIV+ pregnant women (Option B), but a longer course of therapy (Option B+) may reduce vertical transmission in future pregnancies, improve maternal outcomes, and be more cost-effective. We develop a state-transition Markov model, calibrated to HIV demographics and behavior in

Ghana, to evaluate alternative treatment scenarios. The analysis aims to help inform HIV treatment policies in Ghana.

4 - Risk Sensitive Optimal Breast Cancer Diagnosis

Mehmet Ayvaci, Assistant Professor, The University of Texas at

Dallas, Jindal School of Management, 800 West Campbell Road,

Richardson, TX, 75080, United States of America, [email protected], Oguzhan Alagoz, Elizabeth Burnside

We investigate the role of risk preferences in the pursuit of optimal diagnostic decisions after mammography. In particular, we develop a risk-sensitive finitehorizon Markov decision process model to maximize total expected utility of a patient where utility in the Von Neumann-Morgenstern sense is defined over survival duration. We perform sensitivity analyses over the parameters of various utility functions and investigate risk behavior as implied by the clinical practice.

TC03

03- West 102 B- CC

Joint Session DAS/MAS: Military Decision Analysis

Sponsor: Decision Analysis & Military Applications

Sponsored Session

Chair: Gregory Parnell, United States Military Academy, West Point, NY,

10996, United States of America, [email protected]

1 - Air Force Space Command Cyberspace Investment Planning

Gregory Parnell, United States Military Academy, West Point, NY,

10996, United States of America, [email protected]

The Air Force Space Command has used multiple objective decision analysis and optimization for over 15 years to develop their space investment plan.

Recently,the Air Force cyberspace mission was assigned to the command. This paper presents the new cyberspace value model that was developed and implemented for the Air Force Space Command FY15-24 Investment Planning

Process.

314

2 - Sustaining a Multi-attribute Decision Model Over Time

Jim Lowe, Professor, US Air Force Academy, 2354 Fairchild Dr.,

Ste. 6H-238, US Air Force Academy, CO, 80840,

United States of America, [email protected], Gregory Seely

While the process of developing an initial Multi-Attribute Decision Model is well defined, sustaining such a model from year to year becomes an art. As Decision

Makers change, every aspect of the model becomes open for discussion since the new stakeholders have no vested interest in the original model. We present results of a resource allocation model based on 25 years of implementation.

3 - United States Army Brigade Combat Team Model with Demand

Reduction Over Time

Mike Teter, Colorado School of Mines, 1500 Illinois Street, Golden,

CO, 80401, United States of America, [email protected],

Lee Ewing

The US Army is reducing forces over the next few years and requires a model to help senior leader’s to make informed troop level decisions. We present a model to optimally size the United States Army to a certain troop goal while meeting operational, personnel, stationing and budgetary constraints. The model is a mixed integer program with multistage scheduling of demand reduction to a target troop level while lessening the turbulence on soldiers.

4 - New Techniques for Sensitivty Analysis in Multi-objective

Decision Analyisis Models

Jeff Weir, Air Force Institute of Technology, 2950 Hobson Way,

WPAFB, OH, 45433, United States of America, [email protected]

Although a wide variety of sensitivity analysis methods for additive value models appear in the literature, the most commonly used and easily understood is oneway sensitivity analysis. We present two new methodologies, one for one-way and one for n-way sensitivity analysis that address concerns from practice with traditional methods and allow decision makers to quickly investigate how the rank order of alternatives and their values change across the entire feasible region of a weight space.

TC04

04- West 102 C- CC

Data Envelopment Analysis III

Cluster: Data Envelopment Analysis

Invited Session

Chair: Paul Rouse, University of Auckland, Department of Accounting

& Finance, 12 Grafton Road, Auckland, New Zealand, [email protected]

1 - Eco-operational Benchmarking: The Case of Aircraft

Nicole Adler, Hebrew University of Jerusalem, Mount Scopus,

Jerusalem, 91905, Israel, [email protected], Gianmaria Martini,

Nicola Volta

A new, eco-operational DEA directional distance profit function analyses a data set consisting of current engine aircraft combinations from the perspective of airlines, government agencies and the environment. We explore the incentive schemes that may encourage airlines to choose a fleet that is both profitable and green. Furthermore, we identify the trade-off among pollutants and noise for the aircraft and engine manufacturers involved in research and development based on the DEA methodology.

2 - Educational Efficiency and Productivity in Pennsylvanian

Schools

John Ruggiero, University of Dayton, School og Business

Administration, Dayton, OH, 45469-2251, United States of

America, [email protected], Shae Brennan

In this paper, we apply a public sector productivity model to analyze efficiency, environmental costs and productivity of Pennsylvanian schools. Recognizing the importance of exogenous socio-economic factors of production, we apply the model of Johnson and Ruggiero (2012) to decompose overall productivity.

3 - Towards a Typology of DEA Models

Paul Rouse, University of Auckland, Department of

Accounting & Finance, 12 Grafton Road, Auckland, New Zealand, [email protected], Lawrence Seiford

As a discipline evolves there comes a time when there is value in clarifying a common structure and classifying its models into similar types. This aids better communication among researchers and identifies areas that are potentially “over or under” researched. It also facilitates better understanding of common structures and problems that gave rise to a family of models. We describe a typology for DEA models as a first step towards meeting the above objectives.

INFORMS Phoenix – 2012

TC05

05- West 103 A- CC

Statistical Monitoring and Detection

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Justin Chimka, University of Arkansas, 800 W Dickson St,

Fayetteville, United States of America, [email protected]

1 - Implementation of Statistical Process Control on Multiple

Correlated Nonlinear Profiles

Shih-Hsiung Chou, Kansas State University, Manhattan, KS,

66506, United States of America, [email protected], Tzong-Ru Tsai,

Shing I. Chang

Current state-of-the-art methods in profile analysis involve one profile only. This study examines three methods on monitoring multiple correlated nonlinear profiles using B-spline modeling and multivariate EWMA control charts. A twoprofile simulation study is conducted to generate average run lengths for comparison. A real-world data set from a vulcanization process is used to demonstrate the implementation of the proposed methods.

2 - Control Charts for Monitoring Peak Expiratory Flow for

Asthma Diagnosis

Amit Mitra, Professor, Auburn University, College of Business,

419 Lowder Building, Auburn, AL, 36849-5266, United States of

America, [email protected], Jayprakash Patankar

Common measures for asthma diagnosis involve peak expiratory flow or forced expiratory volume in one second as indicators of lung function. In this paper, statistical control charts for monitoring peak expiratory flow, obtained from individual samples, over a time period are developed. The proposed control charts are unique to the health care field.

3 - Economic Model of X-bar Chart with Asymmetric Limits

Triss Ashton, University of North Texas, 1155 Union Circle

#311160, Denton, TX, 76203-5017, United States of America, [email protected], Robert Pavur

This research considers the economic–statistical design of X-bar control charts to minimize the McWilliams’ (1989) cost model using asymmetric limits. Sample averages assume Su (unbounded) Johnson distributions. Optimal parameters are compared with symmetric control limits. Lower costs and smaller sample sizes are reported.

4 - Process Abnormality Detection by Process Time Distribution for

Optimal Planning

Hiroaki Mori, Osaka University, 2-1 Yamadaoka, Suita, Japan, [email protected], Hiroshi Morita

We consider finding the undiscovered defects by using mathematical and statistical model in the steel mill product factory. In this study, we focus on the data of processing time at each process since it may correlate with the level of defectiveness. First, we make a mathematical model including “normal” and

“abnormal” distribution. Then we estimate the defect rate for each process based on the source of defects, and find the optimal process planning to achieve the defect reduction.

TC08

TC07

07- West 104 A- CC

Optimization and Prediction in Data Mining

Sponsor: Data Mining

Sponsored Session

Chair: Chun-An (Joe) Chou, University of Washington, 3900 Stevens

Way, Seattle, WA, United States of America, [email protected]

1 - Exact Solution Approaches for Generalizations to the Order-

Preserving Submatrix Problem

Andrew Trapp, Assistant Professor, Worcester Polytechnic Institute,

School of Business, 100 Institute Rd., Worcester, MA, 01609,

United States of America, [email protected]

The Order Preserving Submatrix (OPSM) problem is an NP-hard data mining problem that, until recently, has eluded exact solution due to its combinatorial nature. Of late, however, there has been some success in solving the OPSM problem exactly using IP-based approaches. Building upon these, we discuss generalizations of the original OPSM problem, such as how to obtain all statistically significant submatrices, and outline solution approaches. Encouraging computational results are provided.

2 - Creating Lists using the Internet

Cynthia Rudin, Assistant Professor, Massachusetts Institute of

Technology, 77 Massachusetts Avenue, Cambridge, MA,

United States of America, [email protected], Benjamin Letham,

Katherine Heller

We want to combine the knowledge of many people (experts) in order to create lists of things that go together, starting from a small seed. This is the same problem that Google Sets was designed to solve. Our algorithm involves a combinatorial search, a large scale clustering algorithm, and an implicit feedback loop.

3 - Multi-task Learning with Heterogeneous Task Relatedness

Yada Zhu, IBM, Yorktown Heights, NY, United States of America, [email protected]

In many applications of multi-task learning, the tasks form multiple groups, and the relatedness between tasks varies. To address this problem, we propose a general framework, which partitions the input features into 2 sets based on their characteristics, and imposes different constraints on their coefficient vectors to accommodate task grouping.

4 - Optimal Multi-voxel Selection in Pattern Classification of

Brain Activity

Chun-An (Joe) Chou, University of Washington, 3900 Stevens

Way, Seattle, WA, United States of America, [email protected],

Kittipat “Bot” Kampa, Art Chaovalitwongse

Recent studies have demonstrated the ability of pattern classification to identify the brain activity with functional MRI (fMRI) technique. A pattern of multiple voxels distributed in the brain conveys discriminative information for cognitive recognition. In our work, we propose a framework to select a set of voxels by introducing a mutual information theory, which is applied to several linear classifiers. We show the experimental result on a multi-class cognition data set.

TC06

06- West 103 B- CC

Agent-based Modeling and Simulation – Overview and Tutorial

Sponsor: Simulation

Sponsored Session

Chair: Charles Macal, Senior Systems Engineer, Argonne National

Laboratory, 9700 S. Cass Ave, DIS - Bldg 221, Argonne, IL, 60439,

United States of America, [email protected]

1 - Agent-based Modeling and Simulation – Overview and Tutorial

Charles Macal, Senior Systems Engineer, Argonne National

Laboratory, 9700 S. Cass Ave., DIS-Bldg 221, Argonne, IL, 60439,

United States of America, [email protected], Michael North

Agent-based modeling and simulation (ABMS) is a new approach to modeling systems comprised of autonomous, interacting agents. Applications are growing rapidly in fields ranging from modeling the stock market to predicting the spread of epidemics. Complex adaptive systems, emergent behavior, and selforganization are a few of the notions from ABMS This tutorial covers the foundations of ABMS, development toolkits and methods, practical aspects, and the relationship of ABMS to conventional OR.

TC08

08- West 104 B- CC

Community-Based Operations Research

Sponsor: Public Programs, Service and Needs

Sponsored Session

Chair: Michael Johnson, Associate Professor, University of

Massachusetts Boston, Dept of Public Policy & Public Affairs,

100 Morrissey Blvd, Boston, MA, 02125, United States of America, [email protected]

1 - Foreclosed Housing Development Strategy Design using

Value-focused Thinking Methods

Michael Johnson, Associate Professor, University of Massachusetts

Boston, Dept. of Public Policy & Public Affairs, 100 Morrissey Blvd,

Boston, MA, 02125, United States of America, [email protected], Rachel Drew, Jeff Keisler,

David Turcotte

Community-based organizations often have trouble designing strategies that link achievement of organizational goals to specific actions. We apply value-focused thinking and decision analysis methods to clarify priorities in order to develop and rank strategies for foreclosed housing redevelopment in three urban CBOs.

Results and sensitivity analyses illustrate differences across missions and service area characteristics.

315

TC09

2 - Optimal Foreclosed Housing Bidding Policies for Community

Development Corporations

Senay Solak, Assistant Professor, University of Massachusetts,

Isenberg School of Management, Amherst, MA, 01003, United

States of America, [email protected], Armagan Bayram

We consider bid/no-bid type decisions for nonprofit community organizations that acquire and redevelop foreclosed properties as part of the societal response to foreclosures. Based on a stochastic dynamic programming formulation, we develop procedures to determine whether a property should be bid on and at what rate. Some numerical analyses are presented based on real-world data.

3 - Decision Modeling for Vacant Lot Acquisition and

Redevelopment

Michael Johnson, Associate Professor, University of Massachusetts

Boston, Dept. of Public Policy & Public Affairs, 100 Morrissey Blvd,

Boston, MA, 02125, United States of America, [email protected], Alma Hallulli, Justin Hollander

Cities facing disinvestment and population loss must decide how to assemble and manage large inventories of vacant lots. We develop decision models for parcel acquisition and redevelopment, with a focus on non-residential uses such as urban farming, stormwater management and community gardens. Our model’s goals include reducing measures of neighborhood distress, supporting environmental mitigation and ensuring political and social feasibility. We apply our model to data from Baltimore, MD.

4 - A Flow Capturing Facility Location Model for Siting Farmers

Markets in an Urban Roadway Network

Jingzi Tan, PhD Student, University of Arizona, 2525 N Los Altos

Ave., Tucson, AZ, 85705, United States of America, [email protected], Wei-Hua Lin

Facility location flow capture model is designed to capture maximum flow based on stochastic customer demands and flexible facility capacity. The behavior of the demand in choosing facilities in the network is incorporated into the model for capacity allocation. Two levels of coverage, the primary and the backup, are also captured in the model.

INFORMS Phoenix – 2012

3 - Diagnostic Assessment of the Borg MOEA for Many-objective

Product Family Design Problems

David Hadka, PhD Candidate Computer Science, The Pennsylvania

State University, 212 Sackett Building, University Park, PA, 16802,

United States of America, [email protected], Timothy Simpson,

Patrick Reed

This study provides a detailed analysis of the Borg Multiobjective Evolutionary

Algorithm (MOEA) on the severely-constrained, ten objective General Aviation

Aircraft (GAA) problem. The Borg MOEA utilizes auto-adaptive search to tailor itself to effectively explore challenging problem spaces. This study benchmarks the Borg MOEA relative to other top MOEAs for the GAA problem using control map analysis and the first application of global variance decomposition on a complex real-world problem.

4 - Locating Diverse and Alternative Solutions in Multiobjective

Civil Infrastructure Design Problems

Brian Piper, North Carolina State University, 2500 Stinson Drive,

Raleigh, NC, United States of America, [email protected],

Ranji Ranjithan

Discovering alternative solutions may aid exploring a solution space to find tradeoffs among competing and conflicting criteria. An evolutionary algorithm for improving decision space diversity and obtaining these alternative solutions is presented using an application to a wastewater management system design problem. Solutions found by the algorithm are analyzed and compared against solutions found by other methods.

TC09

09- West 105 A- CC

Evolutionary Multi-Criterion Optimization - II

Sponsor: Multiple Criteria Decision Making

Sponsored Session

Chair: Patrick Reed, Associate Professor of Civil Engineering, The

Pennsylvania State University, 212 Sackett Building, University Park,

PA, 16802, United States of America, [email protected]

1 - Evolutionary Multiobjective Optimization in Water Resources:

The Past, Present, and Future

Joshua Kollat, Research Associate, The Pennsylvania State

University, 212 Sackett Building, University Park, PA, 16802,

United States of America, [email protected], Jon Herman,

Patrick Reed, Joseph Kasprzyk, David Hadka

This study contributes a rigorous diagnostic assessment of ten state-of-the-art multiobjective evolutionary algorithms (MOEAs) and highlights key advances that the water resources field can exploit to better discover critical system tradeoffs. The applications tested represent the dominant problem classes that have historically shaped the use of MOEAs in water resources. Recommendations are provided for which modern MOEAs should serve as tools and benchmarks in future water resources studies.

2 - Many Objective Visual Analytics: Decision Support for the

Design of Complex Engineered Systems

Matthew Woodruff, PhD Candidate Industrial Engineering, The

Pennsylvania State University, 212 Sackett Building, University

Park, PA, 16802, United States of America, [email protected]

We apply a new many-objective visual analytics (MOVA) approach to the conceptual design of a General Aviation Aircraft product family design problem.

MOVA emphasizes learning through problem reformulation, enabled by visual analytics and many-objective search. Two highly aggregated formulations impose unexpected preferences on solutions, resulting in a myopic view of the problem.

In contrast, a many-objective formulation enables the selection of a wellinformed solution using visual analytics.

TC10

10- West 105 B- CC

Advances in Integer Stochastic Programming and

Testing Solution Quality

Sponsor: Optimization/Stochastic Programming

Sponsored Session

Chair: Honggang Wang, Rutgers University, Piscataway, NJ,

United States of America, [email protected]

1 - Mixed Integer Simulation Optimization for Oil and

Gas Field Development

Honggang Wang, Rutgers University, Piscataway, NJ,

United States of America, [email protected]

In petroleum industry, because well costs can be extremely high, it is essential that wells be drilled in productive locations and operated effectively. Such development problems are greatly complicated in that the geology of the subsurface is highly uncertain. We propose retrospective optimization using dynamic simplex interpolation for mixed integer simulation optimization associated with oil development problems.

2 - Bias and Variance Reduction for Assessing Solution Quality in

Stochastic Programming

Rebecca Stockbridge, University of Arizona,

617 N Santa Rita Ave., Tucson, AZ, 85721, United States of

America, [email protected], Guzin Bayraksan

We present a computational study of bias and variance reduction methods applied to the optimality gap estimators produced by the Multiple Replications Procedure.

Techniques discussed include Latin Hypercube sampling, antithetic variates, and randomized quasi-Monte Carlo for variance reduction, and a probability metrics approach to bias reduction.

3 - An Algorithm for Two-stage Stochastic Programs with Mixed

Integer Recourse

Ted Ralphs, Associate Professor, Lehigh University, 200 West

Packer Avenue, Bethlehem, PA, 18015, United States of America, [email protected], Anahita Hassanzadeh, Menal Guzelsoy

Due to the complex structure of the mixed integer value function, algorithms for two-stage stochastic linear programs with integer recourse have been con?ned to the pure integer case. We propose an algorithm based on (partial) construction of the value function of the second-stage problem. The algorithm is based on characterizing the local minima of the mixed integer recourse value function.

316

INFORMS Phoenix – 2012

TC11

11- West 105 C- CC

Algorithmic Non-linear Optimization

Sponsor: Optimization/Nonlinear Programming

Sponsored Session

Chair: Emre Tokgoz, University of Oklahoma, 202 W. Boyd, Room 124,

Norman, OK, 73071, United States of America, [email protected]

1 - A Modified Potential Reduction Method for Monotone

Complementarity and Convex Programming Problems

Kuo-Ling Huang, PhD Student, Northwestern University,

Evanston, IL, United States of America, [email protected], Sanjay Mehrotra

We present a homogeneous algorithm equipped with a modified potential function for the monotone complementarity problem. We show that this potential function is reduced by at least a constant amount if a scaled Lipschitz condition is satisfied. A practical algorithm based on this potential function is implemented in a software package named iOptimize. Computational results show that iOptimize takes fewer iterations than mature solvers Mosek and Ipopt.

2 - Single-source Capacitated Multi-facility Weber Problem –

An Iterative Two-phase Heuristic Algorithm

S.M.H. Manzour, University of Oklahoma, Oak 2900, Norman,

United States of America, [email protected], S.A. Torabi,

K. Eshghi

Multi Facility Weber (MFWP)Problem entails determining the locations of a predefined number of facilities in a planar space and their related customer allocations. We focus on Single-Source Capacitated MFWP (SSCMFWP). An iterative two-phase heuristic algorithm is put forward. At the phase I, we aim to determine proper locations for facilities, and during the phase II, assignment of customers to these facilities is pursued. The proposed iterative two-phase algorithm produces promising results.

3 - A New Algorithmic Approach to Solve Mixed Variable

Optimization Problems

Emre Tokgoz, University of Oklahoma, 202 W. Boyd, Room 124,

Norman, OK, 73071, United States of America, Emre.Tokgoz-

[email protected]

A well known method to solve mixed variable optimization problems is by using interior point method. In this talk I will be introducing a new method to obtain mixed (i.e. discrete and continuous) convexity and optimization results for mixed variable functions with the corresponding algorithms.

4 - A Hierarchical Algorithm for the Planar Single-facility Location

Routing Problem

S.M.H. Manzour, University of Oklahoma, Oak 2900, Norman,

United States of America, [email protected], S.A. Torabi

Location Routing Problem is an important logistical problem that comprises two of the main logistical drivers namely facility location and vehicle routing. We focus on the Planar Single-Facility LRP with Euclidean distance. A hierarchical heuristic-based method is put forward which continuously takes into account information from the routing results while systematically improving the location using the end points of the obtained routes. The proposed method outperformed the existing approaches.

TC12

12- West 106 A- CC

Joint Session Optimization IP/ICS: Constraint

Programming Methodology and Applications II

Sponsor: Optimization/Integer Programming & Computing Society

Sponsored Session

Chair: Willem-Jan van Hoeve, Carnegie Mellon University,

5000 Forbes Avenue, Pittsburgh, PA, United States of America, [email protected]

1 - Large-scale Power Restoration

Pascal Van Hentenryck, Professor, University of Melbourne,

Melbourne, Australia, [email protected], Carleton Coffrin,

Ben Simon

This paper considers the joint repair and restoration of the electrical power system after significant disruptions. This problem is computationally challenging because, when the goal is to minimize the size of the blackout, it combines a routing and a power restoration component. This paper proposes an hybrid optimization approach based on constraint programming providing high-quality results for infrastructures with more than 24000 components and 1200 damaged items.

TC13

2 - Using Cumulative Propagation to Schedule Resources on

Projects with Release Dates to Pick-up Goods

Thiago Serra, Operations Research Analyst, PETROBRAS, Avenida

Paulista, 901, São Paulo, SP, 01311-100, Brazil, [email protected], Fernando J. M. Marcellino,

Gilberto Nishioka

The development of constraint-based scheduling solvers has been biased towards the satisfaction of the global constraint cumulative. Hence, a successful use of them to tackle many problems depends on leveraging such constraint in the model. In this work, we consider the scheduling of vessels to convey pipes that will connect offshore oil wells to producing units. To guarantee the loading of pipes by activities scheduled after their release, a qualitative property is modeled with such constraint.

3 - New Development in Google OR-Tools

Laurent Perron, Google, 38 Avenue de l’Opéra, Paris, Fr, 75002,

France, [email protected]

Google’s OR-Tools is a set of libraries that includes a constraint programming solver, a wrapper around linear solvers, and some dedicated graph and flow algorithms. Since the initial presentation, development has been very active and lots of new features have been added to the different solvers. We will take advantage of the INFORMS meeting to present these improvements.

4 - MDD Propagation for Disjunctive Scheduling

Willem-Jan van Hoeve, Carnegie Mellon University,

5000 Forbes Avenue, Pittsburgh, PA, United States of America, [email protected], Andre Cire

We present new propagation methods for disjunctive scheduling, based on limited-width Multivalued Decision Diagrams (MDDs). We show how our method can be integrated efficiently with existing propagation algorithms.

Experimental results indicate that the MDD propagation can outperform state-ofthe-art propagators especially when minimizing sequence-dependent setup times, in certain cases by several orders of magnitude.

TC13

13- West 106 B- CC

Solving Hard Problems in Conic Optimization II

Sponsor: Optimization/Linear Programming and Complementarity

Sponsored Session

Chair: Akiko Yoshise, University of Tsukuba, 1-1-1 Tennnoudai,

Tsukuba, 305-8573, Japan, [email protected]

1 - A Perturbed Sums of Squares Theorem and its Applications

Masakazu Muramatsu, The University of Elector-Communications,

1-5-1, Chofugaoka, Chofu-shi, Tokyo, 182-8585, Japan, [email protected], Hayato Waki, Levent Tuncel

We show a property of positive polynomials on a compact set with a small perturbation. This theorem implies that the optimal value of the corresponding

SDP relaxation with sufficiently large relaxation order is bounded from below and above using the size of the perturbation and the number of variables. The SDP relaxation can be of considerably smaller dimensional than Lasserre’s. We present some computational experiments of the SDP relaxation.

2 - Preprocessing and Regularization for Degenerate

Semidefinite Programs

Yuen-Lam Vris Cheung, University of Waterloo, 200 University

Ave., W, Waterloo, Canada, yl[email protected], Simon

Schurr, Henry Wolkowicz

We propose a backward stable preprocessing technique for semidefinite programming (SDP) that applies the Borwein-Wolkowicz facial reduction process to find a rank-revealing rotation of the problem, resulting in a smaller and wellposed problem that can be solved more accurately by standard SDP solvers. We also examine the implications of the minimum number of facial reductions required, on the theoretical and numerical aspects of solving that instance via interior point methods.

3 - A Primal Barrier Function Phase I Algorithm for Nonsymmetric

Conic Optimization Problems

Yasuaki Matsukawa, University of Tsukuba, 1-1-1 Tennnoudai,

Tsukuba, Ib, 305-8573, Japan, [email protected],

Akiko Yoshise

We call the set of positive semidefinite matrices whose elements are nonnegative the doubly nonnegative (DNN) cone. Using its symmetric cone representation, the authors showed that the DNN relaxation gives significantly tight bounds for a class of quadratic assignment problems while the computational time is too long.

Motivated by the observations, we propose a nonsymmetric primal barrier function Phase I algorithm and provide a sufficient condition for finite termination.

317

TC14

TC14

14- West 106 C- CC

Current Issues in Complex Projects

Cluster: Scheduling and Project Management

Invited Session

Chair: Ted Klastorin, Professor, University of Washington,

Foster School of Business, Box 353226, Seattle, WA, 98195-3226,

United States of America, [email protected]

1 - Planning Uncertain Projects under the Threat of

Disruptive Events

Gary Mitchell, Associate Professor, University of Portland, Pamplin

School of Business, Portland, OR, 97203, United States of America, [email protected], Ted Klastorin, Issariya Sirichakwal

We study a project defined by a series of stages where the duration of each stage is determined by the level of allocated resources and random work content described by some distribution. There exists the threat of multiple disruptions; each will stop the project for some random time. We define a model that allocates resources over time to minimize expected total costs (defined by resource costs, overhead/indirect costs, and penalty costs). Optimal policies and implications are discussed.

2 - Managing New Product Development Projects In a

Competitive Market

Issariya Sirichakwal, University of Washington, Foster School of

Business, ISOM Department, Seattle, WA, 98195-3226, United

States of America, [email protected], Hamed Mamani,

Gary Mitchell, Ted Klastorin

In many markets, the first firm to introduce a new product gains a significant benefit. However, reducing the time-to-market typically requires higher development costs that may offset any first-mover advantage. In this paper, we analyze this trade-off problem in a competitive duopoly market from the perspective of a risk neutral firm that wants to maximize the project’s expected total profit. We consider different payoff structures, as well as static and dynamic project scheduling policies.

3 - Optimal Bidding and Investment Strategy in the PPP

Competitive Dialogue

Dennis DeClerck, KBI-Operations Management, Katholieke

Universiteit Leuven, Leuven, Belgium,

[email protected], Erik Demeulemeester

Public private partnerships are cutting edge contractual agreements nowadays.

This research focuses on the competitive bidding procedure with a game theoretical analysis of the influence of the number of competitors, a company’s experience and the government’s compensation schemes. Consequently, a rationally optimal approach for the pre-bidding research strategy and the final offer is proposed.

4 - A Reliability Model for the Stochastic RCPSP

Patricio Lamas, KU Leuven, Naamsestraat 69, Leuven, Belgium,

[email protected], Erik Demeulemeester

In this presentation we consider the stochastic RCPSP where the stochasticity is incurred because either resource availabilities or activity durations are uncertain.

We model the problem as a Chance Constrained Program. The objective is to minimize the project makespan under a predefined reliability level.

TC15

INFORMS Phoenix – 2012

15- West 202- CC

Software Demonstration

Invited Session

1 - SAS - Building and Solving Optimization Models with SAS

Ed Hughes, SAS/OR Product Manager, SAS Institute Inc.,

500 SAS Campus Dr., Cary NC 27513, United States of America, [email protected], Rob Pratt

OPTMODEL from SAS provides a powerful, intuitive algebraic optimization modeling language and unified support for LP, MILP, QP and NLP models. We’ll demonstrate OPTMODEL’s clarity and flexibility in building and solving optimization models, including a review of its newest features. We’ll also explore

OPTMODEL’s integration into a range of SAS data, analytic, and reporting capabilities.

2 - SAS Global Academic Program – Introduction to SAS Rapid

Predictive Modeler

Tom Bohannon, SAS Global Academic Program, SAS Campus Dr.,

Cary NC 27513, United States of America, [email protected]

This presentation is an introduction to SAS Rapid Predictive Modeler, a component of SAS Enterprise Miner. It provides an overview of the product and provides details on using SAS Rapid Predictive Modeler as part of the predictive modeling process and describes the difference between the basic, intermediate and advanced models that are available in SAS Rapid Predictive Modeler.

TC16

16- West 207- CC

Mining Social Media: A Brief Introduction

Cluster: Tutorials

Invited Session

Chair: Pritam Gundecha, Arizona State University, Ira A. Fulton

Schools Engineering, Tempe, AZ, United States of America,

[email protected]

1 - Mining Social Media: A Brief Introduction

Pritam Gundecha, Arizona State University, Ira A. Fulton Schools

Engineering, Tempe, AZ, United States of America,

[email protected], Huan Liu

The pervasive use of social media has generated unprecedented amounts of social data. Social media data is vast, noisy, unstructured, and dynamic in nature, thus novel challenges arise. This chapter reviews the basics of data mining and social media, introduce representative research problems, illustrates the application of data mining to social media using examples, and describes some projects of mining social media for “humanitarian assistance and disaster relief” for realworld applications.

TC17

17- West 208 B- CC

Communication Instruction in Business & Industrial

Engineering: Tools & Recommendations for Your

Own Setting

Sponsor: INFORM-ED

Sponsored Session

Chair: Judith Norback, Dir. of Workplace & Academic Communication,

Stewart School of ISyE,Georgia Tech, 765 Ferst Drive, Atlanta, GA,

30332, United States of America, [email protected]

Co-Chair: Patrick Noonan, Professor in the Practice of Decision & Info.

Analysis, Emory University, 1300 Clifton Road NE, Goizueta Business

School, Atlanta, GA, 30322, United States of America, [email protected]

1 - Problem-driven Communication: Best Practices from the

Consulting Profession

Patrick Noonan, Professor in the Practice of Decision & Info.

Analysis, Emory University, 1300 Clifton Road NE, Goizueta

Business School, Atlanta, GA, 30322, United States of America, [email protected]

Students can learn to communicate effectively and persuasively by adopting methods used by management consultants, and letting the problem content to the talking: relentlessly asking “so what,” creating logic trees (the Pyramid Principle), assembling clear story-lines, and practicing good slide craft. We showcase examples from an experiential-learning project course.

2 - Building and Measuring Business Communication Competency

Karen Eboch, Senior Lecturer, Bowling Green State University,

Department of Management, BAA 3020, Bowling Green, OH,

43537, United States of America, [email protected]

Developing and evaluating communication competency is a key issue for colleges preparing students for career success and business programs dealing with AACSB accreditation. Given today’s business environment, communication skills go beyond basic written and oral proficiency to include more data driven tools and team effectiveness tips. Effective methods of building and measuring these communication competencies across a four year undergraduate business program curriculum will be shared.

3 - Integrating Communication in an Industry Sponsored

Capstone Course

Anita Vila-Parrish, North Carolina State University,

Campus Box 7906, Raleigh, NC, 27695, United States of America, [email protected], Julie S. Ivy, Sarah Egan Warren

This talk will be focused on threading communication tools throughout an industrial engineering capstone course. In this semester-long course, projects are industry sponsored which require students to communicate to different audiences

(internal and external). The course leverages industry standard templates and project management methods to facilitate communication from concept to implementation phase.

318

4 - A 5-Stage Model for Integrating Communication Skills into

Business Curriculums

Lyle Benson, Grant MacEwan University, Edmondton, AB,

Canada, [email protected]

This presentation describes the integration of writing and presenting communication skills into the MacEwan University BCom program and courses via a five-phase model: 1. Institutional Preparation Process, 2. Curriculum

Development Process, 3. Professional Skills and Curriculum Mapping Integration

Process, 4. Evaluation Process, and 5. Faculty Integration Process.

TC18

18- West 208 A- CC

INFORMS Phoenix – 2012

Continuous and Discrete Optimization Software:

State-of-the-art

Sponsor: Optimization/Computational Optimization and Software

Sponsored Session

Chair: Hans Mittelmann, Arizona State University, School of Math and

Stats, Box 871804, Tempe, AZ, 85287-1804, United States of America,

[email protected]

1 - Performance of Commercial and Noncommercial

Optimization Software

Hans Mittelmann, Arizona State University, School of Math and

Stats, Box 871804, Tempe, AZ, 85287-1804,

United States of America, [email protected]

We will report on selected results from our ongoing benchmarking effort. Discrete benchmarks are based on MIPLIB2010 and own selections. Continuous benchmarks include LP, QP, QCQP, SOCP, and SDP problems and solvers.

2 - The SCIP Optimization Suite 3.0 - It’s all in the bag!

Gerald Gamrath, Research Assistant, Zuse-Institut Berlin, Takustr

7, Berlin, 14199, Germany, [email protected], Timo Berthold,

Ambros Gleixner, Stefan Heinz, Marc Pfetsch, Matthias

Miltenberger, Thorsten Koch, Stefan Vigerske, Kati Wolter,

Dieter Weninger, Yuji Shinano, Michael Winkler

We present the SCIP Optimization Suite, a tool for modeling and solving optimization problems. It comes with the modeling language ZIMPL and the LP solver SoPlex and is one of the fastest MIP solvers available in source code.

Additionally, it can solve more optimization problems including non-convex

MINLP and supports branch-and-price. We report on current developments and new features in the 3.0 release, including enhanced MINLP support, exact integer programming, and parallelization.

TC20

3 - Modeling the Effect of Nurse as a Transmitter on Hospital

Acquired Infections

Evrim Didem Gunes, Koc University, Rumeli Feneri Yolu, Istanbul,

Turkey, [email protected], Yasin Arslan, Lerzan Ormeci

We first analyze data from the ICU of a public hospital and show the effect of neighboring beds in probability of acquiring a colonization or infection in the ICU.

Further, we model the ICU acquired infections in the ICU as a Markov Process, accounting for the role of the nurse. We use numerical experiments to investigate policies to reduce infections. Nurse to patient ratio and size of the ICU are major factors that are analyzed.

4 - A Stochastic Model for Epidemics using Spatial Games

Songnian Zhao, PhD Student, Kansas State University, 2037

Durland Hall, Manhattan, KS, 66506, United States of America, [email protected], Chih-Hang Wu, Zhenzhen Shi

We present a stochastic model describing the transmission of infectious diseases, considering the individuals’ spontaneous responses to the epidemics using the spatial game. The study explores the significant factors impacting the dynamics of infectious disease and a case study is given to predict the spread of infectious diseases.

5 - Effectiveness and Cost-Effectiveness of Early Vaccination for a

Human Influenza A (H5N1) Pandemic

David Hutton, University of Michigan, 1415 Washington Heights,

SPH II: M3525, Ann Arbor, MI, 48109, United States of America, [email protected], Nayer Khazeni, Ine Collins

Influenza A (H5N1) is an emerging pandemic threat, yet traditional vaccine development and delivery may take months. We use a mathematical model of pandemic influenza spread to evaluate the health and economic benefits of interventions to speed up the vaccine development and delivery process. Both the health and economic benefits of delivering the vaccine a few months earlier can be substantial. Sensitivity analysis shows conditions that may make speedy vaccine delivery less valuable.

TC19

19- West 211 A- CC

Disease Transmission

Contributed Session

Chair: David Hutton, University of Michigan,

1415 Washington Heights, SPH II: M3525, Ann Arbor, MI, 48109,

United States of America, [email protected]

1 - Calibration of an Agent-based Simulation for Clostridium

Difficile Transmission and Control

James Codella, University of Wisconsin-Madison, 3270 Mechanical

Engineering, 1513 University Avenue, Madison, WI, 53706, United

States of America, [email protected],

Oguzhan Alagoz, Nasia Safdar

Clostridium difficile infection (CDI) affects 500,000 Americans every year, and causes nearly 20,000 deaths annually. We propose an agent-based simulation to model the transmission dynamics and strategies to control CDI. Furthermore, we describe methods to calibrate unknown parameters and validate assumptions for an agent-based simulation.

2 - Probabilistic Risk Assessment of Influenza Control Strategies using an Agent-based Model

Elnaz Karimi, Graduate Research Assistant, Concordia University,

Industrial Engineering, 1455 de Maisonneuve Blvd W, EV-S3.330,

Montreal, QC, H3G 1M8, Canada, [email protected],

Ali Akgunduz, Ketra Schmitt

We introduce an agent-based model to simulate the behavior of each individual in a university population during a seasonal influenza outbreak considering both pharmaceutical and non-pharmaceutical interventions. We then conduct a costeffectiveness analysis to find the best combination of interventions to control the spread of influenza in the target population. Finally we evaluate the results of this model within the context of probabilistic risk assessment (PRA).

TC20

20- West 211 B- CC

Information in Health Care

Contributed Session

Chair: Subodha Kumar, Mays Business School, Texas A&M University,

College Station, TX, 77843, United States of America, [email protected]

1 - Dynamics of Change: Institutional Logics of Adoption and

Use Health IT

Anthony James Baroody, Lecturer, Rochester Institute of

Technology, E. Philip Saunders College of Business, 107 Lomb

Memorial Drive, Rochester, NY, 14623, United States of America, [email protected], Sean Hansen

This study extends earlier research analyzing institutional logics at play in the adoption of electronic health records (EHRs) within the U.S. healthcare system.

Institutional logics are combined with system dynamics to examine the interplay of multiple institutional logics upon EHR adoption. Complementary and conflicting institutional logics are examined in a systems perspective to evaluate implications for the emergence of an IT-intensive U.S. healthcare system.

2 - Analyzing Process Outcomes and Care Coordination with Digitization

Yi-Chin Lin, Carnegie Mellon University, 5000 Forbes Ave.,

Hamburg Hall, Pittsburgh, PA, 15213-3890, United States of

America, [email protected], Rema Padman,

Lakshmanan Krishnamurti

Although broadly advocated, the impact of Health IT remains unclear. This study analyzes data from clinician surveys and electronic medical records to evaluate the impact of digitization of paper-based, individual pain plans on process outcomes and care coordination in management of children with sickle cell pain presenting in emergency department.

3 - Electronic Medical Record (EMR) and Physician

Career Satisfaction

Fei Li, PhD, University of Southern California, USC, Los Angeles,

CA, 90089, United States of America, [email protected]

This study uses survey to assess the association of physician career satisfaction and potentially significant factors of EMR system and implementation process. Based on Technology of Acceptance and Use of Technology (UTAUT) model and successful cases studies, the factors are developed in four constructs: perceived usefulness of EMR, perceived ease of use, social influence and facilitating conditions during implementation process. Factor analysis and logistic regression are performed.

319

TC21

INFORMS Phoenix – 2012

4 - Physicians interactions with Electronic Health Records in

Primary Care

Onur Asan, UW-Madison, 1513 University Avenue, Madison, WI,

53705, United States of America, [email protected],

Enid Montague

It is essential to design technologies and systems that promote appropriate interactions between physicians and patients. This study explored how physicians interact with EHRs to understand the qualities of the interaction between the physician and the EHR that may contribute to positive physician-patient interactions.Future research on this topic and design strategies for effective health information technology in primary care are also discussed.

5 - Healthcare Information Exchange and Interoperability:

A Game-Theoretic Approach

Subodha Kumar, Mays Business School, Texas A&M University,

College Station, TX, 77843, United States of America, [email protected], Arun Sen, Emre Demirezen

US government reimburses healthcare information exchanges (HIEs) and healthcare providers if they join HIE networks. HIEs also offer other value-added services for additional fees. Benefits and costs of providers depend on the service levels they receive. We discuss when providers should join HIE, and the level of service they should request. On the other hand, HIEs maximize their profits considering the cost of different services and deciding on the prices of these service offerings.

TC21

21- West 212 A- CC

Disaster and Evacuation Management

Sponsor: Optimization/Networks

Sponsored Session

Chair: Chrysafis Vogiatzis, University of Florida, Weil 303, P.O. Box

116595, Gainesville, FL, 32611-6595, United States of America, [email protected]

1 - Evacuation Planning for Livestock in a Case of a Nuclear Power

Plant Accident

Ruriko Yoshida, University of Kentucky, 325D MDS building,

Lexington, KY, 40506, United States of America, [email protected], Ines Aviles-Spadoni, Chrysafis Vogiatzis,

Shigeki Imamoto

Development of a mathematical formulation of an evacuation plan for livestock around a nuclear power plant is essential for farmers to lessen financial, emotional impacts from losing their animals. So we propose two mathematical models for evacuation plan for livestock via integer programming (IP) over networks. Due to NP-completeness of IP, we propose a heuristic algorithm for our problem. We end with simulation results/experiments with the map near the

Fukushima Daiichi nuclear power plants.

2 - Clustering Techniques and Hybrid Algorithms in Evacuation

Management

Chrysafis Vogiatzis, University of Florida, Weil 303, P.O. Box

116595, Gainesville, FL, 32611-6595, United States of America, [email protected], Panos Pardalos, Jose L. Walteros

Evacuation management is of utmost importance for the well-being of every modern society. In this presentation, we emphasize on large-scale evacuation of metropolitan areas. The problem, being extremely large, is almost impossible to solve without employing a decomposition of the network at hand. We present an approach to effectively decompose the problem in a series of smaller evacuation problems and, then, effectively combine the optimal solutions provided by the approximate subproblems.

3 - Optimal Allocation of Renewable and Non-Renewable

Resources in Disasters

Shengbin Wang, Rutgers University, 1 Washington Park, Newark,

NJ, 07102, United States of America, [email protected],

Lei Lei

We consider an allocation problem over a two-stage supply chain network with renewable and non-renewable resources, which is encountered from the practice of providing medical service to people in the affected area by a disaster. This situation is modeled as a MIP to coordinate both resources to minimize the total tardiness. We analyze the problem properties and based on that we provide a heuristic algorithm to solve it efficiently.

4 - Multi-commodity Stochastic Network Design:

Without Conservation-of-flow on the Design Variables

Xin Wang, PhD Student, Lancaster University Management

School, House 172, Graduate College, Lancaster University,

Lancaster, LA2 0PJ, United Kingdom, [email protected],

Stein Wallace

We present a stochastic, time-dependent, capacitated, multi-commodity service network design problem in which periodic, cyclic schedules are built for a number of vehicles. We aim to find and illustrate the underlying structures of solutions based on stochastic demand by studying the optimal solutions to the stochastic network design problem and comparing them with deterministic counterparts, and to provide insight into developing a heuristic for large problems.

TC22

22- West 212 B- CC

Computational Integer Programming

Sponsor: Computing Society

Sponsored Session

Chair: Yan Xu, Analytical Solutions Manager, SAS Institute Inc., SAS

Campus Drive, Cary, SC, United States of America, [email protected]

1 - Recent Advances in the Xpress MIP solver

Oliver Bastert, FICO, Starley Way, Birmingham, United Kingdom,

[email protected], Michael Perregaard

We will present some of the recent developments in the Xpress MIP solver, with particular emphasis on heuristics. Modellers continually push the boundaries on the size of problems that can be solved and is often satisfied with a solution that is

“good enough”. This talk will focus on the developments in Xpress to address such problems.

2 - Gurobi Presolve

Robert Bixby, CEO and President, Gurobi Optimization Inc.,

3733-1 Westheimer Rd., Houston, TX, United States of America, [email protected], Zonghao Gu, Ed Rothberg

Presolve refers to that step before solution of an optimization model in which an attempt is made to simplify a model. For LPs these simplifications remove

“extraneous” parts of the model, but for MIPs the changes can be more fundamental, producing a strengthened model, and this strengthening can make the difference between a model being solvable and hopeless. We will examine

Gurobi presolve, including the effects of some of the more interesting new reductions.

3 - Developments in SAS MILP Solver

Amar Narisetty, Operations Research Specialist III, SAS Institute,

100 SAS Campus Drive, Cary, NC, 27513, United States of

America, [email protected]

The SAS MILP solver implements a branch-and-cut algorithm for solving mixed integer linear programs. In this talk, we give an overview of the recent improvements that are incorporated in the SAS MILP solver. These improvements are in presolve, LP, cuts, heuristics and a new option for decomposable MIPs. We present computational results to demonstrate the effectiveness of these improvements.

4 - Recent Improvements in IBM ILOG CPLEX Optimization Studio

Laszlo Ladanyi, CPLEX R&D, IBM, P.O. Box 800, Lakeville, CT,

06039, United States of America, [email protected]

We present new features that have been added to IBM ILOG CPLEX Optimization

Studio and give benchmarking results that demonstrate the performance improvements in CPLEX 12.5.

320

TC23

23- West 212 C- CC

Innovation in Analytics Award: Semi-finalist

Presentations III

Sponsor: Analytics

Sponsored Session

Chair: Michael Gorman, University of Dayton, Dayton, OH,

United States of America, [email protected]

1 - Insurance Agency Productivity, Efficiency and Prospecting

Mark Grabau, IBM, United States of America, Elizabeth Riczko

Insurance companies desire a fact-based, decision-making process for managing their agency distribution channel such that it is productive and profitable. We present an approach that combines Data Envelopment Analysis, Predictive

Modeling, Collaborative Filtering, and Reporting to objectively meet that need.

Results are discussed as well as areas for future research and on-going improvements.

2 - A System for Optimizing Advance Acquisition of Right-of-Way for Highways at TxDOT

Ron Hagquist, Texas Transportation Institute (TTI),

125 E. 11th Street, Austin, TX, 78701, United States of America,

Dong Hun “Don” Kang, Carlos M. Chang-Albitres, Paul E. Krugler,

Sergiy Butenko, Richard M. Feldman, Reza Seyedshohadaie

Acquiring highway right-of-way well in advance of actual construction can avoid substantial escalation of the cost of purchasing parcels. However, this money could also be used to advance other aspects of existing projects underway. A decision-support system combining simulation and optimization was developed for field use analyzing this classic tradeoff situation, with an optimal minimumcost solution depending on the parameters of the specific situation.

3 - Sustainable Investment Decision Analysis: Integrated

Techno-economic Analysis for a Biofuel Initiative

Scott Mongeau, Principal Consultant, Nyenrode Business

University

An integrated modeling and assessment technique known as techno-economic analysis is emerging as a promising method to address the inherent complexity associated with sustainable energy project decision management. Via a practical case involving a bioethanol R&D project for a biomaterials conglomerate, key considerations related to techno-economic decision analysis are examined. A unique process-oriented analysis approach was composed and applied to guide decision making for the firm which resulted in the launch of a collaborative partnership venture to deploy an experimental cellulosic biofuel plant. The integrated approach, involving NPV modeling, Monte Carlo analysis, and Real

Options Analysis (ROA), is discussed in terms of methodological, technical, and practitioner considerations.

4 - Data-driven Appointment Scheduling in the Presence of No-shows

Michele Samorani, Alberta School of Business, University of

Alberta, Edmonton, AB, Canada, [email protected],

Linda LaGanga

We consider the problem of scheduling outpatient appointments in the presence of no-shows. Predictive analytics is used to forecast the show outcome of appointment requests, which are optimally scheduled given this prediction.

Descriptive analytics is used to interpret the solutions of the scheduling problem in order to derive a heuristic policy, which, depending on the show rate and on the prediction quality, outperforms the reigning scheduling policy, open access.

Prescriptive analytics is used to identify the conditions under which one should adopt overbooking or same-day scheduling. We validate our findings on the data set of a large mental health center.

TC24

24- West 213 A- CC

Game-Theoretic Applications in Healthcare

Sponsor: Health Applications Society

Sponsored Session

Chair: Murat Kurt, Assistant Professor, University at Buffalo, SUNY,

University at Buffalo, North Campus, 415 Bell Hall, Buffalo, NY, 14260,

United States of America, [email protected]

1 - Optimal Design of the Annual Influenza Vaccine with

Autonomous Manufacturer

Osman Ozaltin, Assistant Professor, University of Waterloo, 200

University Avenue West, CPH 3671, Waterloo, ON, N2L 3G1,

Canada, [email protected], Andrew Schaefer, Oleg Prokopyev

Seasonal influenza is a major public health concern, and the first line of defense is the flu shot. We model the annual influenza strain selection problem as a bilevel stochastic mixed-integer program. Calibrated over publicly available data, our model determines the optimal flu shot composition and production in a stochastic and dynamic environment. We also analyze the effects of yield uncertainty, price, and production cost on the flu shot production.

2 - Coordinating Vaccine Markets: Operational Issues and

Network Effects

Elodie Adida, University of California at Riverside, School of

Business Administration, Riverside, United States of America, [email protected], Hamed Mamani, Debabrata Dey

Vaccines are the most effective means for preventing infectious diseases. However, negative network externalities on the consumption side and operational issues

(such as yield uncertainty) on the supply side do not provide the incentives required to reach the socially optimal vaccine coverage. We investigate how a central policy-maker can induce a socially optimal coverage through the use of a two-part subsidy scheme.

INFORMS Phoenix – 2012

TC25

3 - DMA-POMDP: A Multi-agent Framework for Decentralized

Control and its Application to Outbreak Response

Emine Yaylali, North Carolina State University, 375 Daniels Hall,

Raleigh, NC, 27695, United States of America, [email protected],

Julie S. Ivy

We present a multi-agent partially observable Markov decision process framework

(DMA-POMDP) for decentralized control of multiple agents under uncertainty.

DMA-POMDP can be applied to problems where agents have partial information about both other agents and system state and communication between agents is explicitly formulated. We apply the DMA-POMDP framework to a public health setting where local and state health departments seek to optimize the timing of response under a disease outbreak.

4 - Equilibrium Stability vs Payoff Efficiency in Prearranged Paired

Kidney Exchanges

Murat Kurt, Assistant Professor, University at Buffalo, SUNY,

University at Buffalo, North Campus, 415 Bell Hall, Buffalo, NY,

14260, United States of America, [email protected],

Mark Roberts, Utku Unver, Andrew Schaefer

Paired kidney exchanges (PKE) alleviate the shortage in the supply of kidneys for transplantation. We model patients’ transplant timing decisions in a PKE as a stochastic game and address the trade-off between the payoff efficiency and the stability of the equilibria against patients’ joint deviations. To characterize a socially efficient stable equilibrium, we formulate an MILP representation of the resulting equilibrium conditions. We present computational results using clinical data.

TC25

25- West 213 B- CC

Healthcare Modelling

Sponsor: Health Applications Society

Sponsored Session

Chair: Michael Carter, University of Toronto, 5 King’s College Road,

Toronto, ON, M5S 3G8, Canada, [email protected]

1 - Cross-sector Patient Flow Model: Towards a Conceptual

Framework for Patient Flow Management

Ali Vahit Esensoy, PhD Candidate, University of Toronto,

5 King’s College Road, Toronto, ON, M5S 3G8, Canada, [email protected], Michael Carter

Developing policies from a system-wide perspective is becoming increasingly important as health authorities promote interventions, which have multi-sector implications. We present a system dynamics model of patient flows between health system sectors, which relates policy levers, performance measures and flow characteristics, effectively providing a conceptual framework for patient flow through the healthcare system. Examples are given on the model’s applications in policy making in Ontario.

2 - A Generalized Perioperative Simulation as a Decision and

Diagnostic Tool

Daphne Sniekers, PhD Candidate, University of Toronto,

5 King’s College Road, Toronto, ON, M5S 3G8, Canada, [email protected], Michael Carter

A generalised, simulation-based perioperative decision support tool has been designed to inform tactical level decision making. The primary intention of the tool was to evaluate various decisions regarding the effect of scheduling and availability of perioperative resources on patient flow. Through application at six hospitals, it was found that the tool is not only useful for decision making, but also to diagnose various efficiency issues that lead to poor outcomes.

3 - A Decision Support Tool for Hip and Knee Osteoarthritis Health

Service Delivery

Sonia Vanderby, Assistant Professor, University of Saskatchewan,

57 Campus Drive, Saskatoon, SK, S7N 5A9, Canada, [email protected], Deborah Marshall, Robert Lee,

Paul Rogers, Michael Carter, Tom Noseworthy

The rising prevalence of osteoarthritis is increasing the burden on already strained healthcare resources. Alberta Health Services managers lack tools to help inform difficult resource and process decisions. We present a system dynamics model developed as a decision support tool for the hip and knee osteoarthritis system.

Simulating patients from disease onset through both medical and surgical management, the model provides insight into the resources needed to meet current and future demand.

321

TC26

INFORMS Phoenix – 2012

4 - Optimization-based Tools for Residency Scheduling

Amy Cohn, University of Michigan, Ann Arbor, MI,

United States of America, [email protected], Brian Jordan,

Young-Chae Hong

The scheduling of medical residents is a complex combinatorial optimization problem that is often solved manually by Chief Residents. We present a MIPbased approach that dramatically reduces scheduling time while eliminating errors that are ubiquitous in practice. More importantly, the resulting schedules are better designed to ensure safe and quality patient care and better educational experiences for the residents.

TC26

26- North 221 A- CC

Operations Economics: Dynamic Models

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Gabriel Weintraub, Columbia Business School,

3022 Broadway, New York, NY, 10027, United States of America, [email protected]

1 - A Supplier’s Response to Auditing and Incentives for Social and

Environmental Performance

Terry Taylor, University of California, Berkeley, Haas School of

Business, Berkeley, CA, 94720, United States of America, [email protected], Erica Plambeck, Qinqin Zhang

Firms audit their suppliers’ labor conditions and environmental performance, and even provide financial incentives for such performance. Unfortunately, suppliers often respond with deception - by hiding noncompliance. We identify conditions under which higher auditing effort or financial incentives backfire - causing the supplier to invest less in compliance. We also identify conditions under which a buyer should commit to invite a third party to audit is supplier and to publicize violations.

2 - Promoting Competition in Repeated Procurement Auctions

Gabriel Weintraub, Columbia Business School, 3022 Broadway,

New York, NY, 10027, United States of America, [email protected], Sachin Adlakha

Most of the literature in auctions assumes a one shot interaction between the auctioneer and the bidders. However, in many real-world procurement settings there are many repeated interactions and suppliers make entry, exit, and other decisions over time. We introduce a novel dynamic model of repeated procurement auctions with these features. We consider commonly used mechanisms to enhance competition (subsidies, split-awards, set-asides) and study which ones lead to a lower procurement cost.

3 - Dynamic Price Signaling Strategies

Laurens Debo, University of Chicago, Chicago, IL,

United States of America, [email protected]

In this talk, I will discuss dynamic pricing strategies of a seller that is privately informed about the quality of an asset.

4 - Social Norms in Queues

Gad Allon, Northwestern University, 2001 Sheridan Rd, Evanston,

IL, United States of America, [email protected]

In many service settings customers have adopted self-enforcing priority rules.

While in some cases, all customers are served according to the order in which they arrive, cutting the line is possible in other systems. We provide conditions under which these intrinsic priorities may emerge. Our results suggest that when priority rules are not centrally managed, they depend on arrival and service rates, customer heterogeneity and patience.

TC27

27- North 221 B- CC

Interface of Operations and Finance

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Guoming Lai, University of Texas at Austin,

1 University Station, Austin, TN, United States of America, [email protected]

Co-Chair: Song Alex YangAssistant Professor, London Business School,

Sussex Place, London, NW1 4SA, United Kingdom, [email protected]

1 - The Impacts of Inventory Sharing on Commodity Procurement

Seung-Jae Park, Universityof Texas, Austin, 2110 Speedway Stop

B6500, Austin, TX, 78712, United States of America, Seung-

[email protected]texas.edu, Guoming Lai, Sridhar Seshari

322

We consider two firms that use a common commodity input to satisfy stochastic demands in a multi-period setting. The firms procure the commodity as well as sell excess inventory through either the spot or the forward market. The firms also share the commodity. We characterize the firms’ equilibrium procurement policies. We show inventory sharing is always beneficial. We numerically find that the option to share inventory may either increase or decrease the firms’ optimal inventory levels.

2 - Which Operational Parameters Matter for the Valuation of

Electricity Storage?

Yangfang Zhou, PhD Candidate, Carnegie Mellon University,

5000 Forbes Ave., Pittsburgh, United States of America, [email protected], Nicola Secomandi, Jay Apt,

Alan Scheller-Wolf, Stephen Smith

We study how the market valuations of three grid-scale battery electricity storage technologies depend on their operational characteristics, such as the number of executable cycles and the magnitude of the charging and discharging rates. We discuss which of these characteristics are most valuable for each technology. We also investigate how the frequency of decision making affects our findings.

3 - Improving Operational Competitiveness through Bankruptcy

Song Alex Yang, Assistant Professor, London Business School,

Sussex Place, London, NW1 4SA, United Kingdom, [email protected], Rodney Parker, John Birge

We study how a distressed firm can use bankruptcy reorganization to improve operational competitiveness and how this strategy influences its competitor, supplier, and creditors. A successful reorganization allows the bankrupt firm to lower its operational cost and that this outcome improves the efficiency of the chain overall. However, in the long run, easy reorganization benefits the supplier and the non-distressed competitor but may hurts the distressed firm.

4 - Going Out of Business: Applying Management Science to Retail

Chain Liquidation

Nathan Craig, Doctoral Candidate, Harvard Business School,

Soldiers Field Rd, Boston, MA, 02163, United States of America, [email protected], Ananth Raman

Retail chain liquidation is the time-constrained sale of inventory located at multiple retail locations. The retail industry depends on chain liquidation to allow investors to recover funds from failed ventures and to enable managers of going concerns to divest stores in efforts to improve performance. This paper introduces techniques for increasing the efficiency of retail chain liquidation. Further, this paper demonstrates the performance of these techniques across multiple liquidations.

TC28

28- North 221 C- CC

Global Health Operations

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Jeremie Gallien, London Business School, Regent’s Park,

London, NW14SA, United Kingdom, [email protected]

1 - Optimal Food Aid Policies for Minimizing Childhood

Malnutrition in Sub-Saharan Africa

Lawrence Wein, Professor, Stanford University, Graduate School of

Business, Stanford, CA, United States of America, [email protected], Yan Yang, Jan Van den Broeck

Using longitudinal data from 5167 children in the Democratic Republic of Congo, we model the evolution of height and weight from birth to age five as a bivariate integrated Ornstein-Uhlenbeck process. We find food aid policies that minimize the disability-adjusted life-years (DALYs) related to wasting and stunting, and compare our policies to the currently recommended policies.

2 - Improving Laboratory Networks for Early Infant HIV Diagnosis in Sub-Saharan Africa

Jeremie Gallien, London Business School, Regent’s Park, London,

NW14SA, United Kingdom, [email protected], Sarang Deo,

Jonas Oddur Jonasson

Many sub-Saharan countries experience high mortality rates among HIV+ infants.

An important cause is the long delay to obtain diagnosis due to ineffective systems for transporting and analyzing blood samples and communicating results.

Using a representative dataset combining information from several countries, we develop a queueing network model to estimate the potential public health impact of operational interventions such as clinic assignment optimization and segmentation of samples.

3 - Network Externality in Allocation of Point-of-care (POC)

Diagnostic Tests in Developing Countries

Sarang Deo, Assistant Professor, Indian School of Business,

Gachibowli, Hyderabad, 500032, India, [email protected],

Milind Sohoni

POC diagnostic devices aim to eliminate diagnostic delays and improve patient retention in developing countries. We develop an optimization model for the allocation of a limited number of devices over a network of clinics. We study the impact of network externality – allocation to one clinic changes the delay experienced at other clinics – which is not accounted for in simple thumb rules used in practice. We apply our results to an infant HIV diagnosis program in a sub-Saharan country.

TC29

INFORMS Phoenix – 2012

29- North 222 A- CC

Appointment Sequencing and Scheduling

Sponsor: Manufacturing & Service Oper Mgmt/ Healthcare

Operations/SIG

Sponsored Session

Chair: Rachel Chen, University of California at Davis, Graduate School of Management, Davis, United States of America, [email protected]

1 - Appointment Scheduling under Patient Preference and

No-show Behavior

Nan Liu, Columbia University, 600 W 168th St., Room 603, New

York, NY, 10032, United States of America, [email protected],

Jacob Feldman, Huseyin Topaloglu, Serhan Ziya

We develop a model to schedule patient appointments over time. On each day, we decide which appointment days to make available for the patients who raise appointment requests on the current day. Our model explicitly considers patient preference and no-show behavior. We derive the optimal static policies and show that they have a performance guarantee which improves as demand increases.

We also develop dynamic policies which utilize the current state information and improve over the static ones.

2 - Appointment Scheduling with Nonlinear Waiting Costs and No-shows

Michael Pinedo, Professor and Chair, New York University, Stern

School of Business, 44 West 4th Street, New York, NY, 10012,

United States of America, [email protected], Christos

Zacharias, Kangbok Lee

Consider an appointment scheduling problem with a fixed number of slots and a number of customers to be assigned to the slots. Each customer has a weight and a given probability of not showing up. The waiting cost function of the customers is nonlinear. We establish conditions under which certain priority rules are optimal for certain segments of the schedule.

3 - Sequencing Surgeries for Operating Rooms with

Inventory Approximations

Jiawei Zhang, Stern School of Business, New York University,

New York, NY, United States of America, [email protected],

Ying Rong, Ho-Yin Mak

Managing surgery appointments for operating rooms (ORs) is a challenging task due to randomness in surgery durations. We propose new heuristics for the problem of surgery sequencing, i.e., determining the order in which a list of surgeries should be performed in an OR. The fundamental idea behind the development of these heuristics is the structural connection between OR appointment scheduling and stochastic inventory control in serial supply chains.

4 - Appointment Sequencing: Is the Smaller-Variance-First

Rule Optimal?

Zhichao Zheng, PhD Candidate, National University of Singapore,

1 Business Link, Biz 2 Building, #B2-03, Singapore, 117592,

Singapore, [email protected], Qingxia Kong, Chung-Yee Lee,

Chung Piaw Teo

We study the design of appointment system to minimize the expected waiting time of patients and overtime of the doctor when patients’ service durations are random. It is widely conjectured that sequencing patients with smaller variance in service time first might be optimal. We challenge this conjecture and show that it may not be optimal in general. We investigate the complexity of the problem and propose a robust planning framework. Some important insights are drawn from numerical analysis.

TC30

TC30

30- North 222 B- CC

Resource Allocation

Contributed Session

Chair: Alexander Galenko, Operations Research Analyst, AgileAssets,

Inc, 3144 Bee Caves Rd, Austin, TX, 78746, United States of America, [email protected]

1 - Fair Allocation of Scarce Resources after a Tornado – Case

Study Norman Area

Mohammad Nikoukar, Graduate Student, The University of

Oklahoma, 202 W Boyd Street, Carson Engineering Center,

Room 35, Norman, OK, 73019, United States of America, [email protected]

This article focuses on the allocation of scarce resources after a natural disaster, case study a tornado in Norman, Oklahoma area. Since the scarce recourses are vary in different situations, it is really critical to assign them correctly. Fair allocation approach can be helpful to consider both rescuers and rescuees. The most important factor in rescuees perspective, is the number of people that be helped but rescuers like to experience uniform work load distribution.

2 - Advanced Aggregate Planning Decisions with Optimal Balance between Demand and Supply

Kris Lieckens, Dr., KU Leuven, Naamsestraat 69, Leuven, 3000,

Belgium, [email protected], Nico Vandaele

An advanced multi-product, multi-routing system is presented to support aggregate planning decisions regarding product-mix, production volume and capacity. Given demand in multiple periods, the model decides on the optimal periodic quantity to be assigned to each alternative, while taking into account relevant costs and variabilities in demand and production. The methodology is a queueing network with estimates for lead time distribution and service level.

Managerial insights are shown.

3 - A Resource Allocation Model for Assigning Boats to Stations in the United States Coast Guard

Christie Nelson, Rutgers University, 640 Bartholomew Rd.,

Piscataway, NJ, 08854, United States of America, [email protected], Matthew Oster, Endre Boros,

Todd Aikins, Thomas Sharkey, William M. Pottenger, Paul Kantor,

Fred Roberts, Chad Conway, Kevin Hanson, Patrick Ball,

Kim Babcock, James Wojtowicz

We present a resource allocation model for assigning boats of varying types to stations within the United States Coast Guard (USCG), so as to meet mission-hour requirements and budget constraints. We discuss an initial integer programming formulation which can be solved to near-optimality on instances which include a small subset of stations (e.g. a single district). We then demonstrate a decomposition algorithm for handling the entire USCG data set. Numerical test results will be presented.

4 - Cross-training Decisions for a Multi-Shift Workforce

Allocation Problem

Aditya Gandhi, Graduate Research Assistant, Rochester Institute of

Technology, 1 Lomb Memorial Drive, Rochester, NY, 14623,

United States of America, [email protected], Scott Grasman,

Shrikant Jarugumilli

This work develops a cross-training framework and applies it to the multi-shift workforce allocation problem. Two different approaches are being evaluated and discussed. The first apporach is that of building an integrated model which adds cross-training parameters and scheduling constraints to the problem. The second approach is to develop different cross-training policies and evaluate them to obtain a policy which can best reduce overtime allocations and operational gaps.

5 - Asset Management Models: Model Size Reduction in the

Context of Pavement Management System

Alexander Galenko, Operations Research Analyst, AgileAssets, Inc,

3144 Bee Caves Rd, Austin, TX, 78746, United States of America, [email protected], Tonya Scheinberg

The lack of resources for economic agents that followed the Great Recession resulted in a substantial increase of asset management literature. Typical topics of that literature are optimization models or optimization algorithms. Surprisingly, the techniques that can be used to reduce model size are not well covered. The authors present a number of ways that can be used to reduce model size while producing near optimal solutions in the context of Pavement Management

System

323

TC32

INFORMS Phoenix – 2012

TC31

31- North 222 C- CC

Empirical Analysis in Services and Retail

Sponsor: Manufacturing & Service Oper Mgmt/Service Operations

Sponsored Session

Chair: Antonio Moreno-Garcia, Kellogg School of Management,

2001 Sheridan Road, Evanston, 60208, United States of America, [email protected]

1 - Override the Revenue Management System? Dynamic Price

Rivalry in the Airline Industry

Jun Li, University of Pennsylvania, Wharton School, Philadelphia

PA, United State of America, [email protected],

Serguei Netessine

Most airline revenue management systems in use do not factor in competition.

Fare managers may override the system based on how competitors price. Using high-frequent carrier-flight level fare and booking information in selected

European routes, we study airline price rivalry. How airlines react to their rivals’ price adjustments? We also disentangle the impact of competition and inventory on revenue management controls.

2 - Traffic Arrivals Into Retail Stores - Distributional Assumptions and Why They Matter?

Vidya Mani, Pennsylvania State University, State College, PA,

United States of America, [email protected], Jayashankar M.

Swaminathan, Saravanan Kesavan

We study the distribution of customer arrivals to retail stores in a large retail chain. Most analytical work on customer arrivals assumes a Poisson distribution for planning purposes as it is analytically tractable. Using hourly traffic data from a major retailer, we identify the underlying distribution of incoming traffic and explore the implications of using an incorrect arrival distribution by studying its impact on store labor planning and profitability.

3 - Empirical Analysis of Inventory Display in Retail

Antonio Moreno-Garcia, Kellogg School of Management,

2001 Sheridan Road, Evanston, 60208, United States of America, [email protected], Santiago Gallino

Using a combination of observational data and a field experiment, we explore the consequences of inventory display strategies in retail environments across different categories.

4 - Matching Supply with Demand in an Outpatient Clinic

Diwas KC, Emory University, 1300 Clifton Road, Atlanta, GA,

30322, United States of America, [email protected],

Nikolay Osadchiy

We examine patient scheduling at an outpatient clinic, where a significant number of scheduled appointments end up as no-shows. In this research, we explore factors that lead to the mismatch between demand and supply, and propose potential solutions to help the clinic manage its capacity utilization.

TC32

32- North 223- CC

Strategic Operations and Marketing Decisions in a

Supply Chain

Sponsor: Manufacturing & Service Oper Mgmt/Supply Chain

Sponsored Session

Chair: Muge Yayla-Kullu, Asst. Prof. Operations Mgmt., RPI Lally

School of Mgmt, 110 8th St, Troy, NY, 12180, United States of America,

[email protected]

1 - Managing Inventory with Limited History of Intermittent

Demand

Alp Akcay, PhD Candidate, Carnegie Mellon University,

5000 Forbes Ave., Pittsburgh, PA, 15213, United States of America, [email protected], Sridhar Tayur, Bahar Biller

Assuming an intermittent demand process with unknown distribution parameters, we study the problem of estimating inventory targets from limited amount of historical demand data. The resulting inventory-target estimates minimize the average expected cost that is attributable to misestimating the optimal inventory target. We also consider the correlation between demand size and the number of inter-demand periods, and propose an integrated estimationoptimization approach to solve the problem.

2 - Entry Deterrence Strategies Facing Capacitated Competition

Huaqing (Mike) Wang, University of Miami, Department of

Management, Miami, FL, 33146, United States of America, [email protected], Haresh Gurnani, Murat Erkoc

A monopolistic incumbent firm faces entry of a capacity-constrained firm. As function of the entrant’s capacity, the incumbent firm can adopt different strategies to try to deter or accommodate the entrant. We also show that when entry is inevitable, the capacitated entrant (incumbent) can get higher profit in

Stackelberg competition as follower (leader) than in Nash Competition. We also consider a case in which the incumbent also invests in changing consumers’ product preference.

3 - Retailer Buy-back and Supply Chain Performance

Jun Zhang, The University of Texas at Dallas,

800 W. Campbell Rd., Richardson, TX, United States of America, [email protected], Suresh Sethi, Owen Ma

We study the retailer buy-back program and its impact on supply chain performance. We show that the retailer buy-back program, when properly designed, can improve the performance of all supply chain members.

4 - Global Supply Chain Networks

Nitish Jain, INSEAD, 1 Ayer Rajah Avenue, Singapore, 138676,

Singapore, [email protected], Karan Girotra,

Serguei Netessine

We study the evolution of global supply chain networks of publicly traded U.S.

retail and wholesale firms. We construct these networks using a novel data-set that is compiled by processing over 10 million import transactions by U.S. firms from global suppliers.

5 - Collaboration Among Supply Chain Partners: Better Outcomes or More Conflict?

Muge Yayla-Kullu, Asst. Prof. Operations Mgmt., RPI Lally School of Mgmt., 110 8th St., Troy, NY, 12180, United States of America,

[email protected]

The emerging area of supply chain alliances studies how firms work together for mutual benefit. In this paper, we investigate benefits of such an alliance for collaborative procurement. We take heterogeneity of the customer base into account when the products of firms are differentiated. In such a supply chain network, we find that joint consideration of sourcing, pricing, and demand generation objectives may result in non-trivial outcomes.

TC33

33- North 224 A- CC

Inventory Management

Contributed Session

Chair: Xue Lu, PhD Student, London Business School, Sussex Place,

London, NW1 4SA, United Kingdom, [email protected]

1 - Economic Lot Sizing: The Capacity Reservation Model

Xi Li, HKUST, Clear Water Bay, Kowloon, Hong Kong,

Hong Kong-PRC, [email protected], Chung-Yee Lee

Capacity reservation contracts allow a consumer to purchase up to a certain capacity at a unit price lower than that of the spot market, while the consumer’s excess orders are realized at the spot price. In this paper, we consider a lot sizing problem where the consumer places orders following a capacity reservation contract. In particular, we study the general problem and the polynomial time solvable special cases of the problem, and propose corresponding algorithms for them.

2 - A Combined Transportation-Inventory Problem for a

Multi-Product Probabilistic Demand Environment

Uladzimir Rubasheuski, PhD Scholar, H¯gskolen i Molde, Britvegen

2, Molde, 6410, Norway, [email protected],

Johan Oppen, David L. Woodruff

This paper is devoted to the analysis of the tradeoff between transportation and inventory costs for a multiple products flow on a single link. A linear model has been developed to minimize total costs of transportation and inventory handling by determining the optimal combination of safety factor, replenishment cycle length and delivered quantity for each replenishment.

3 - Optimal Policy and Time Consistency for Multi-stage Inventory

Models with Free Distributions

Linwei Xin, PhD. Student, Georgia Tech, 765 Ferst Drive NW, ISyE,

Georgia Tech, Atlanta, GA, 30332, United States of America, [email protected], Alexander Shapiro

In a seminal work on inventory optimization, H. Scarf found an explicit solution to a certain minimax problem with moment constraints and piecewise-linear cost function, when the cost function has only two pieces. In this work, we extend his result to more general cost functions, and discuss several applications in multistage inventory models.

324

4 - The Value of Yield Information in a Periodic Review

Inventory System

Marcus Dettenbach, Universtiy of Cologne, Albertus Magnus Platz,

Department of Supply Chain Management, Cologne, 50923,

Germany, [email protected], Ulrich Thonemann

We consider the number of defective units in an order to be random. The information on the number of defective units is available at any time. This setting is compared to a setting in which the yield can only be observed upon arrival of an order. We solve a dynamic program for both settings. To solve larger problems, we develop close-to-optimal heuristics with short run times. We conduct various numerical experiments and identify dependencies for the value of real time yield information.

5 - A Horizon Decomposition Algorithm for the Multi-level Capacity

Constrained Lot Sizing Problem

Xue Lu, Phd Student, London Business School, Sussex Place,

London, NW1 4SA, United Kingdom, [email protected],

Zeger Degraeve

Multi-level capacity constrained lot sizing (MLCLS) has high application value, but solving the problem to optimality remains daunting. We propose a new

Horizon Decomposition algorithm for the echelon stock formulation of MLCLS.

We decompose the time horizon into contiguous short time horizons. Our algorithm provides a tighter lower bound than all the existing model reformulations, and also CPLEX by 30% within the same computing time.

TC34

34- North 224 B- CC

Innovation Contests

Cluster: New Product Development

Invited Session

Chair: Anant Mishra, Assistant Professor, George Mason University,

4400 University Drive, Fairfax, VA, United States of America, [email protected]

1 - Submission Transparency in Idea Generation

Joel Wooten, University of Pennsylvania, The Wharton School,

Philadelphia, PA, United States of America, [email protected], Karl Ulrich

In innovation tournaments, administrators face a variety of decisions that impact the course of the contest. We examine the effect of submission transparency by comparing blind and unblind contests using innovation tournament field experiments. We control submission visibility and show individual-level differences in idea quality and uniqueness from the tournament participants - the solvers.

2 - Cognitive Response of Workers to Competition

Michael Menietti, Harvard Business School, Boston, MA, United

States of America, [email protected], Karim Lakhani,

Kevin Boudreau, Constance Helfat

Rank-order tournaments are widely employed to incentivize workers and elicit effort. We examine software developers competing in rank-order tournaments.

We find performance falls among developers in response to the number of competitors and independently from the presence of superstar competitors.

Though those affected still work on solutions, fewer solutions are submitted and more solutions that are submitted contain errors.

3 - Prize Amount and Entry Behavior in Innovation Contests

Anant Mishra, Assistant Professor, George Mason University,

4400 University Drive, Fairfax, VA, United States of America, [email protected], Cheryl Druehl, Jesse Bockstedt

Innovation contests are being widely used by firms to generate creative ideas and identify solutions to complex problems. Using a large panel dataset of unblind innovation contests from a popular online logo-design platform, we examine the interplay or prize amount and contestant’s prior winning experience on their entry behavior in a contest.

4 - The Impact of Monetary and Non-monetary Incentives in

Platform-based Innovation Contests

Annika Mueller, Harvard University, Boston, MA,

United States of America, [email protected],

Karim Lakhani, Kevin Boudreau

In this study we examine the effect of pecuniary and non-pecuniary incentives on the participation, effort and productivity of individuals on an open crowdsourcing contest platform. Specifically, we conduct a unique field-experiment in which subjects are randomly allocated to nine treatment groups to investigate the causal impact of various combinations and levels of money, status and job market signalling as prizes on the participation and performance of coders in an online programming contest.

INFORMS Phoenix – 2012

TC36

TC35

35- North 225 A- CC

Optimization Models in Revenue Management

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: Huseyin Topaloglu, Cornell University, 223 Rhodes Hall,

Ithaca, NY, United States of America, [email protected]

1 - Assortment Optimization under General Choice

Srikanth Jagabathula, New York University, Kaufman

Management Center, 44 W 4th Street, Rm 8-74 KMC,

New York, NY, 10012-1126, United States of America, [email protected], Vivek Farias, Devavrat Shah

We consider the classic decision problem of static assortment optimization: find an offer set that maximizes revenue subject to a constraint on its size. We assume access to only revenue estimates for different offer sets. Finding the optimal solution in general requires exhaustive search over offer sets. We propose an approximation algorithm with far fewer calls to revenue function than exhaustive search requires. We also establish theoretical and exhaustive empirical analysis.

2 - Assortment Optimization under Variants of the Nested Logit

Choice Model

Huseyin Topaloglu, Cornell University, 223 Rhodes Hall,

Ithaca, NY, United States of America, [email protected],

Guillermo Gallego, James Davis

We consider assortment optimization problems under several variants of the nested logit model. Under the assumption that the no-purchase is not available in each nest and the so-called nest dissimilarity parameters are between zero and one, we show that we can compute the optimal assortment in polynomial time.

The problem is NP-hard when we relax either one of these assumptions. We explore approximation methods for the NP-hard versions of the problem.

3 - Joint Bidding and Pricing Models

Karti Puranam, Assistant Professor, LaSalle University,

1900 Olney Avenue, Philadelphia, PA, United States of America, [email protected], Michael Katehakis

We present a class of models where a firm procures items by participating in auctions and it decides at what price to sell them. We discuss conditions for the existence of stationary optimal policies of simple structure for the present value and the average-reward optimal policies.

4 - Self-adjusting Price Control for Network Revenue Management

Stefanus Jasin, Assistant Professor, Ross School of Business,

University of Michigan, 701 Tappan St., Ann Arbor, MI,

United States of America, [email protected]

We provide a very simple improvement for the well-known static price control of

Gallego and Van Ryzin (1997). Our approach requires only linear updating of target sales rate and does NOT require any re-optimization at all. This suffices to improve the O(T^0.5) regret bound of static control to O(log T). For a certain class of problems, we show how to do semi-decentralized control. Finally, we briefly discuss whether it is possible to update the price of only SOME of the products instead of all.

TC36

36- North 225 B- CC

Revenue Optimization and Related MDP

Methodologies I

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: Michael Katehakis, Department Chair and Professor, Rutgers

University, Rutgers Business School, Newark, NJ, 07102, United States of America, [email protected]

1 - Resource Allocation in Stochastic Contests with Exponential

Completion Times

Pelin Canbolat, Dr., Technion, Haifa, Israel, [email protected], Uriel Rothblum, Boaz Golany

We study a class of games where players compete to complete a task first by deciding on their resource allocation. Completion times are assumed to be stochastically independent and exponentially distributed with rate linear in the amount of resource allocated. Only the first player to complete the task earns a reward. We explore the mean-variance tradeoffs based on explicit representations of payoff probability distributions, means, and variances for any feasible resource allocation.

325

TC37

2 - Average-Cost Markov Decision Processes with Weakly

Continuous Transition Probabilities

Eugene A Feinberg, Professor, Stony Brook University,

Stony Brook, NY, United States of America, [email protected], Pavlo Kasyanov, Nina Zadoianchuk

This talk presents broad sufficient conditions for the existence of stationary optimal policies for discounted and average-cost Markov Decision Processes with

Borel state and action sets and with weakly continuous transition probabilities.

The one-step cost functions may be unbounded, and the action sets may be noncompact. The results allow multiple applications including applications to pricing problems.

3 - Optimal Adaptive Policies for Sequential Allocation

Problems – Revisited

Apostolos Burnetas, Professor, University of Athens, Department of Mathematics, Panepistemiopolis, Athens, 15784, Greece, [email protected], Michael Katehakis

Consider the problem of sequential sampling from a finite number of statistical populations under incomplete information. In a paper in 1996 we showed how to construct index-based policies with uniformly maximum convergence rate for the case in which the distributions of outcomes from each population depend on a vector of unknown parameters. That work is generalized in several directions regarding dependence of observations. Applications in pricing and data mining are presented.

4 - Estimating Willingness-to-Pay and Willingness-to-Sell

Distributions from Incomplete Negotiation Data

A. Serdar Simsek, PhD Candidate, Columbia Business School,

3022 Broadway, Uris Hall, 4L, New York, NY, 10027, United States of America, [email protected], Garrett Van Ryzin,

Robert Phillips

We develop a method for using the Expectation-Maximization (EM) algorithm to estimate willingness-to-pay (WTP) and willingness-to-sell (WTS) distributions using incomplete negotiation data. The data are incomplete because the final price is recorded only if trade actually takes place and WTP and WTS values are not observed. The results are promising in terms of the accuracy of the computed estimates and when used to predict take-up on both synthetic and real data sets.

5 - Inventory-based Dynamic Pricing with Costly Price Adjustment

Sridhar Seshari, Professor, University of Texas at Austin, McCombs

School of Business, Austin, TX, 78712, United States of America,

[email protected], Wen Chen, Qi Feng

We study a long-run inventory planning problem in which the retailer can replenish inventory and change price anytime. We establish that it is optimal to change the price from low to high in each replenishment cycle, the optimal orderup-to level may decrease when the ordering cost increases, and fewer customers are served when the unit cost of procurement increases. Additionally, we provide efficient algorithms to compute the stocking and pricing policies.

TC37

37- North 226 A- CC

Energy Infrastructure Modeling

Sponsor: Location Analysis

Sponsored Session

Chair: Michael Lim, Assistant Professor, University of Illinois,

1206 S. 6th St., Champaign, IL, 61820, United States of America, [email protected]

1 - Competition Among Food, Energy and Environment:

Sustainable Biofuel Supply Chain Design

Xin Wang, University of Illinois at Urbana-Champaign, 111 E

Healey St Apt 202, Champaign, Il, 61820, United States of

America, [email protected], Michael Lim, Yanfeng Ouyang

Some attributes the recent significant raise in food price to rapid expansion of biofuel industry. We propose a continuous biofuel supply chain model to (i) capture the impact of biofuel production on the food market and environment; and (ii) draw managerial insights for the biofuel firms as well as the government.

2 - Renewable Energy Supply Chain Design under Oligopolistic

Competition and Uncertainty

Zhaomiao Guo, University of California-Davis, Department Civil &

Environment Eng., Davis, CA, 95616, United States of America, [email protected], Yueyue Fan

In this paper, we model the renewable energy supply chain network problem involving oligopolistic firms, who are competing for resources, transmission capacity, and demand market. The decision variables include the production quantities and capacities of energy infrastructures. The non-corporative market behavior is captured by equilibrium conditions. Uncertainties in supply and demand are addressed by stochastic programming techniques.

INFORMS Phoenix – 2012

3 - Facility Location Design for Large Scale Energy Storage

Devices

Yong Liang, PhD Candidate, University of California-Berkeley,

Etcheverry Hall 1117, Berkeley, CA, United States of America, [email protected], Z. Max Shen

Recently there has been growing interest in large-scale electricity storage facilities.

In this talk, we construct a model to formulate the problem of selecting locations and sizes of the storage facilities. The objective is to minimize the upfront construction costs plus the long-run operation costs. We also conduct analysis on minimizing the worst cast costs when there are uncertainties associated with the locational marginal pricing and local demands.

4 - Flexible and Convenient Car Sharing System Design and Operations

Guangrui MA, PhD Student, Hong Kong University of Science and

Technology, Room 3208, Department of IELM, Clear Water Bay,

Kowloon, Hong Kong-PRC, [email protected]k, Ho-Yin Mak

Currently, car sharing becomes a popular solution to keep the balance between individual driving and sustainability. But it is criticized for its non-convenience and inflexibility from re-positioning coordination difficulty, which may lower the cars’ utilization. The company Car2go announced its car sharing service overcoming these problems. We try to capture the business model, analyze the critical issues, compare with the traditional ways, and show how it influences the whole social warfare.

TC38

38- North 226 B- CC

Supply Chain Services: Models and Applications-II

Sponsor: Service Science

Sponsored Session

Chair: M. Ali Ulku, Assistant Professor, School of Management and

Leadership, Capital University, 1 College and Main, Columbus, OH,

43209, United States of America, [email protected]

1 - Incentive Contracts Between 4PLs and 3PLs: An Application of the Principal-agents Model

Qin Zhu, City University of Hong Kong, Tat Chee Avenue,

Kowloon, Hong Kong SAR, Kowloon, Hong Kong-PRC, [email protected], Y.K., Richard Fung

This paper applies the principal-agent theory to the logistics industry, studying the horizontal cooperation between 4PLs and 3PLs through incentive contracts, where the compensation is a combination of a fixed fee and a bonus based on the performances of the 3PLs, in terms of the service delivery rate and customer satisfaction level.

2 - Managing Risks of Information Asymmetry and Demand

Uncertainty in Assembly Supply Chains

Xiang Fang, Assistant Professor, University of Wisconsin-

Milwaukee, 3202 N Maryland Avenue, Milwaukee, WI, 53211,

United States of America, [email protected], Jun Ru,

Yunzeng Wang

We study a decentralized assembly supply chain in which an assembler (she) assembles a set of components, each produced by a different supplier (he), into a final product to satisfy an uncertain market demand. Each supplier holds private cost information to himself, for which the assembler only has a subjetive estimate.

The assembler aims to find an optimal menu of contracts in order to maximize her own expected profit and induce each supplier to truthfully reveal his private cost information.

3 - A Decision Support System for Shipment Decisions in a Supply

Network with Price Dependent Demands

Asli Sencer, Assoc. Prof., Bogazici University, Hisar Campus,

Department of Management, Information Systems, Bebek,

Istanbul, 34342, Turkey, [email protected], Bertan Badur

In this study a supply network is considered with multiple distributors and retailers. The demands at the retailers are dependent on the selling prices at the destination points. A transportation model with price dependent demands is considered to optimally determine the shipment quantities as well as the prices at the destination points. The model is integrated with a database and user interfaces are developed to form a decision support system (DSS).

326

4 - Thanks, But No Thanks?: Impact of Donor Behavior on the

Humanitarian Logistics Chain

Stephanie Gray Wilson, Assistant Professor, Captial University,

Department of Psychology, Columbus, OH, United States of

America, [email protected], Kathryn M. Bell, M. Ali Ulku

Disaster response requires prompt and efficient organization and delivery of supplies and services in the Humanitarian Logistics Chain (HLC). Solicited and unsolicited donations are frequently provided by various, often uncoordinated, funding sources to assist with disaster relief. This paper explores the impact of different types of funding sources and donations on the HLC and offers a framework of how donor behavior impacts operations. We provide a research outlook for this emerging topic.

TC39

39- North 226 C- CC

Topics in Operations Management I

Contributed Session

INFORMS Phoenix – 2012

Chair: Mohamed Ismail, Assistant Professor, University of Regina,

3737 Wascana Parkway, Rgina, SK, S4S0A2, Canada, [email protected]

1 - Bus Product Configuration System Based on Mass

Customization

Zhiduan Xu, Professor, Xiamen University, Room 606,Building

Jiageng 1, 422 Siming South Road, Xiamen, FJ, 361005, China, [email protected]

In this paper, a product configuration system is built in order to fulfill different customized orders efficiently and effectively in the Bus industry by creating product families and product platforms. Moreover, the process from the order to the manufacturing is reconstructed based on the system.

2 - A Methodology for Assessing Rework Risks in Projects

Theodore Glickman, Professor of Decision Sciences, George

Washington University, 2201 G Street, NW, Duques Hall,

Washington, DC, 20052, United States of America, [email protected], Luis Novoa

A methodology for assessing the risks of rework is introduced. A novel approach for representing the rework associated with inspection outcomes as an augmentation of the original activity network is proposed. Risk profiles for project time and cost are developed and displayed in terms of exceedance probabilities.

3 - Lead-time Quotation When Customers are Sensitive to Reputation

Susan Slotnick, Department of Operations and Supply Chain

Management, Cleveland State University, 1860 E. 18th Street,

Cleveland, OH, 44115, United States of America, [email protected]

This model considers how reputation should influence lead-time optimization.

Qualitative monotonicity results are not obtainable, so a computational study explores the relationships between shop status, order size, reputation, market characteristics and the lead-time decision. Regression analysis sheds light on these relationships and suggests three heuristics, which provide near-optimal solutions with relatively short running times.

4 - Progressive Modeling: Towards a New Systems

Optimization Paradigm

Mohamed Ismail, Assistant Professor, University of Regina,

3737 Wascana Parkway, Rgina, SK, S4S0A2, Canada, [email protected]

Progressive Modeling (PM) is a multidisciplinary forward looking systems optimization approach that finds pragmatic solutions for many complex/large scale industrial problems. PM brings many innovations to problem analytics, modeling, and solution algorithms. This paper introduces many PM principles and demonstrates a handful of PM implementations in different manufacturing environments.

5 - The Multilevel Rationing Policy for Repairable Systems

Pedram Sahba, Research Assistant, University of Toronto, 5 King’s

College Road, Toronto, ON, Canada, [email protected],

Baris Balcioglu

A repairable system consisting of m manufacturing plants with finite number of machines and a single repair shop is considered in this work. We propose the multilevel rationing policy(MR) for dispatching the spare parts. The MR policy prioritizes classes, and stops serving a class from inventory if the inventory level is below the inventory threshold of that class. We compare the performance of the

MR policy with the optimal dynamic policy and the existing dispatching policies in the literature.

TC40

TC40

40- North 227 A- CC

Joint Session ENRE-Env & Sustainability/Energy:

Green Biofuel Supply Chain Management

Sponsor: Energy, Natural Res & the Envi/ Environment and

Sustainability & Energy, Natural Res & the Environment/Energy

Sponsored Session

Chair: Yongxi Huang, Assistant Professor, Clemson University,

314 Lowry Hall, Clemson, SC, 29634, United States of America, [email protected]

1 - An Optimization Model for the Facility Sizing and Location of

Thermochemical Biofuel Production

Guiping Hu, Iowa State University, 3004 Black Engineering, Ames,

IA, 50010, United States of America, [email protected], Yihua Li

This paper investigates the fast pyrolysis – upgrading – refining pathway, with corn stover being the feedstock and gasoline being the end product. Mixed integer linear programming (MILP) model is formulated to optimize the fast pyrolysis and upgrade facility location and capacity to minimize total production cost. The economic feasibility of building a new bio-refinery in Iowa is analyzed with the comparison to the existing bio-refinery in Louisiana.

2 - Effects of Air Pollutant Emissions Control on the Cost and

Spatial Distribution of Biorefineries

Colin Murphy, Graduate Student Researcher, University of

California Davis, 2028 Academic Surge, 1 Shields Ave., Davis, CA,

95616, United States of America, [email protected],

Nathan Parker

UC Davis has developed, a spatially explicit technoeconomic model of large-scale biofuel production in the U.S.Biofuel production may lead to significant air pollutant emissions from agricultural and biorefinery activity. Current research adds costs of additional air pollution control devices to biorefineries in PM and

Ozone nonattainment areas, to determine whether the costs of compliance significantly alter the final fuel costs or spatial distribution of biorefineries.

3 - Optimization of Sustainable Biofuel Supply Chain under Uncertainty

Yongxi Huang, Assistant Professor, Clemson University,

314 Lowry Hall, Clemson, SC, 29634, United States of America, [email protected], Fei Xie

This study presents a multi-objective stochastic programming approach for cellulosic biofuel supply chain design under uncertainty of conversion technology efficiency. A compromise programming model is formulated to integrate economic competitiveness and environmental quality into system planning and operations. The findings of case study of California show that the technology efficiency can alter the system configurations, however through smart system modeling, the effects can be minimized.

4 - Algorithms for Economic Lot Sizing Models with multi-Mode

Replenishments and Carbon Considerations

Gokce Palak, Mississippi State University, P.O. Box 9542,

Mississippi State, MS, 39762, United States of America, [email protected], Sandra Eksioglu

We investigate extensions of economic lot sizing models under carbon emission considerations as it applies to the transportation planning decisions. We propose algorithms for these extensions that consider single fixed charge and multiple setups cost structures. Our objective is to show how the changes in the supplier and transportation mode selection choices would contribute to reduction of carbon emissions for alternative carbon regulatory mechanisms.

327

TC41

INFORMS Phoenix – 2012

TC41

41- North 227 B- CC

Forestry Session IV

Sponsor: Energy, Natural Res & the Environment/Forestry

Sponsored Session

Chair: Marc McDill, Associate Professor, Pennsylvania State School of

Forest Resources, 310 Forest Resources Building, University Park, PA,

16802, United States of America, [email protected]

1 - Construction of Routes on Harvest Areas for Harvesters and

Forwarders using Detailed GIS Information

Patrik Flisberg, Norwegian School of Economics, Helleveien 30,

Bergen, Norway, [email protected], Sima Mohtashami,

P. Jönsson, Mikael Ronnqvist

Today detailed information on harvest areas including tree locations, terrain condition, altitude and soil conditions can be used to optimize the trail outlay for harvesters and forwarders. We propose a solution method to construct both primary and secondary trails for both machine types. We present results from case studies in Sweden.

2 - Coordination Mechanisms in Forest Value Chains

Sophie D’Amours, Université Laval, Department of Mechanical

Engineering, Pavillon Adrien Pouliot, Quebec, QC, G1K 7P4,

Canada, [email protected], Mikael Ronnqvist

Logistic planning is often based on decoupled planning of several independent problems. In these cases, there is a need to introduce coordination mechanisms to find the best solution for the entire chain. We explore a number of planning problems and describe efficient coordination mechanisms based on different decomposition strategies.

3 - Planning under Uncertainty

Patrik Flisberg, Norwegian School of Economics, Helleveien 30,

Bergen, Norway, [email protected], Mikael Frisk,

Mikael Ronnqvist

There are many applications in forestry with uncertain data. We explore solution methods and strategies for forest transportation and harvesting when some data is uncertain. Results from case studies are presented.

4 - Routing with Queuing

Bertil Liden, SkogForsk, Uppsala, Sweden, [email protected], Patrik Flisberg, Mikael Ronnqvist

Routing of trucks is often done with the assumption that routes are independent.

This can often lead to extensive queuing at larger mills and it is important to consider this aspect. We propose a solution method which adds on queuing as an important constraint. We present results from case studies in Sweden.

TC42

42- North 227 C- CC

Transit System Modeling

Sponsor: Transportation Science & Logistics/ Urban Transportation

Sponsored Session

Chair: Yueyue Fan, Associate Professor, University of California,

Department of Civil & Environmental Eng., Davis, CA, 95616,

United States of America, [email protected]

1 - Continuum Approximation Model for Transit System Design under Variable Demand

Seyed Mohammad Nourbakhsh, PhD Student, University of

Illinois at Urbana-Champaign, B156 Newmark Civil Engineering

Laborator, Urbana, IL, 61801-2352, United States of America, [email protected], Yanfeng Ouyang

This research proposes a continuum approximation model for transit network design considering variable passenger demand. The model determines the optimal structure of the transit network as well as the bus stop spacing and headway.

Numerical experiments illustrate the effective network design for the system under various demand assumptions.

2 - Optimal Operations, Management, and Network Design for

Personal Rapid Transit Systems

Ryan Smith, Graduate Research Assistant, University of Illinois at

Urbana-Champaign, Newmark Civil Engineering Laboratory,

Urbana, Il, 61801-2352, United States of America, [email protected], Yanfeng Ouyang

This paper presents an integrated personal rapid transit planning model that uses a genetic algorithm and an embedded closed queuing network model to optimize the pod fleet size, station location, the number of berths at each station, and the strategy for empty pod reallocation. The model formulates agency investment and user cost as functions of design characteristics.

3 - Optimizing Urban Rail Timetable under Time-dependent

Demand and Oversaturated Conditions

Xuesong Zhou, University of Utah, 110 Central Campus Dr, MCE

2000, Salt Lake City, UT, 84112, United States of America, [email protected], Huimin Niu

This talk focuses on optimizing a passenger train timetable in a heavily congested urban rail corridor. Based on time-dependent, origin-to-destination trip records from an automatic fare collection system, a nonlinear optimization model is developed to capture the overall passenger delay, subject to the available fleet.

Based on cumulative input-output diagrams, several solution algorithms are presented.

TC43

43- North 228 A- CC

Unit Train Planning and Scheduling

Sponsor: Railway Applications

Sponsored Session

Chair: Erdem Eskigun, CSX, 500 Water Str, Jacksonville, FL, 32202,

United States of America, [email protected]

1 - Coal Monthly Reservations Planning

Ece Icyuz, University of Florida, 369 Maguire Village, Apt 4,

Gainesville, FL, 32603, [email protected], Jean-Philippe Richard,

Erdem Eskigun, Dharma Acharya

Each month, CSX must determine how to best allocate its resources to satisfy the largest possible number of coal reservations. This problem can be formulated as a

MIP of gigantic size that cannot be solved with off-the-shelf software. To obtain solutions, we develop a heuristic that produces, in the first phase, an aggregate solution that is later refined in the second phase. We provide some computational results evaluating the speed and quality of solutions using real application data.

2 - Unit Train Generator (UTG) – A Statistical Approach to Forecast

Unit Trains

Viraj Karnik, Manager Operations Research, Norfolk Southern

Corporation, 1200 Peachtree St NE, MS 12-117, Atlanta, GA,

30309, United States of America, [email protected],

Ilya Lavrik, Tonya Woods

Unit trains form a significant part of Norfolk Southern’s transportation business.

In this talk we present Unit Train Generator (UTG) – an automated and integrated tool developed by NS Operations Research for transportation planners. UTG, based on monthly unit traffic volume forecast, creates train schedules, i.e. it determines train origin – destination, train route, crew districts, frequency, number of cars, arrival/departure/dwell times.

3 - Freight Railway Operator Timetabling and Engine Routing

Lukas Bach, PhD. Student, Aarhus University, Fuglesangs alle 4,

8000, Aarhus V, Denmark, [email protected], Michel Gendreau,

Sanne W¯hlk

We consider timetable design at a European freight railway operator. This is done by choosing the time of service for demands among discrete points of service within a time-window. The objective in the model is to minimize cost while adhering to constraints regarding infrastructure usage, demand coverage and engine availability. The model is solved by a column generation scheme where feasible engine routings are designed in a label setting algorithm with timedependent cost and service times.

4 - Unit Train ETA Forecasting

Erdem Eskigun, CSX, 500 Water Str, Jacksonville, FL, 32202,

United States of America, [email protected],

Dharma Acharya

We present an ETA forecasting tool developed to better estimate the arrival time of CSX unit trains to their destinations. The tool not only considers the standard and historical transit and dwell time data but also the real time delay information to have better forecasts. The delays considered might be due to curfews, derailments, no crew, no engine, and customer shut-downs. ETA forecasts are refreshed at every real time event like arrival, departure, crew call, and delay at route stations.

328

INFORMS Phoenix – 2012

TC44

44- North 228 B- CC

Supply Chain, Practice and Empirics: Inventory

Contributed Session

Chair: Kai Hoberg, Associate Professor of Supply Chain and Operations

Strategy, Kühne Logistics University, Brooktorkai 20, Hamburg,

Germany, [email protected]

1 - The Effect of Excess Inventory on Operating Performance

Kevin Hendricks, Professor, School of Business & Economics,

Wilfrid Laurier University, Waterloo, On, N2L 3C5, Canada, [email protected], Vinod Singhal

This talk will present empirical evidence on the effect of excess inventory on return on assets, return on sales and sales over assets. The results are based on a sample of nearly 800 publicly traded companies that announce that they have excess inventory. The median decline in ROA ranges from 2.17% to 3.49% over a three-year period around the year of the excess inventory announcement. The median decline in ROS (SOA) ranges from 1.58% to 2.77% (6.13% to 6.91%).

2 - Industrial Vending and VMI: Usage, Challenges, and Successes

John Kros, Professor of Marketing and Supply Chain

Management, East Carolina University, 3205 Harold Bate,

Greenville, NC, 27858, United States of America, [email protected]

Almost everyone has used a vending machine at some point in time to get a soda or snack. Vending machines aren’t confined to just the cafeteria. Industrial vending machines provide solutions to many different Supply Chain challenges.

This research is to help gain a better understanding of increased efficiency & cost saving opportunities offered by industrial vending machines & vendor managed inventory (VMI). Questions as “Who are the players? What are the costs & benefits?” are addressed.

3 - Inventory Management under Financial Distress:

An Empirical Analysis

Kai Hoberg, Associate Professor of Supply Chain and Operations

Strategy, Kühne Logistics University, Brooktorkai 20, Hamburg,

Germany, [email protected], Mario Pesch

Financial distress is a threat to the survival of many firms. Manufacturers typically have high amounts of capital tied up in inventories that could be transformed to cash in distressed situations. We leverage empirical data for 5,489 US firms in the

1994-2007 period to identify their actions under financial distress. We find evidence that a sample of 198 distressed manufacturers significantly reduce their inventory by an average 9.4% to ensure liquidity.

4 - Improving Agility of Forest Supply Chain through Adding

Flexibility in the Wood Procurement Phase

Shuva Gautam, Université Laval, Département des Sciences du

Bois et de la Forêt, Faculté de Foresterie, de Géographie et de

Géomatique, 2405, Rue de la Terrasse, Pavillon Abitibi-Price,

Quebec, QC, G1W2L7, Canada, [email protected],

Daniel Beaudoin, Luc LeBel

Wood procurement planning models for the forest supply chain accept predefined silvicultural decisions made at an upper hierarchical level. Flexibility in such decisions at the operational level where demand data is more accurate should improve the supply chain agility. However, making silvicultural decisions at the operational level requires the planners to anticipate and ensure long-term sustainability of their decision. We propose a mixed integer programming model to addresses these issues.

TC45

45- North 229 A- CC

Panel Discussion on Women in Engineering:

Publishing, Recruitment, and Retention

Sponsor: Women in OR/MS

Sponsored Session

Chair: Laura McLay, Virginia Commonwealth University, Statistics &

Operations Research, 1015 Floyd Ave, Box 843083, Richmond, VA,

23284, United States of America, [email protected]

1 - Panel Discussion on Women in Engineering: Publishing,

Recruitment, and Retention

Moderator: Laura McLay, Virginia Commonwealth University,

Statistics & Operations Research, 1015 Floyd Ave., Box 843083,

Richmond, VA, 23284, United States of America, [email protected], Panelists: Mary Anderson-Rowland,

Robyn McKay

The panelists will discuss issues for women in OR/MS as they relate to publishing, recruitment, and retention of women. The panelists are listed as co-authors in this talk.

TC47

TC46

46- North 229 B- CC

Game Theory II

Contributed Session

Chair: Aadhaar Chaturvedi, Assistant Professor, University of Namur

(FUNDP), Rempart de la Vierge 8, Bureau 406, B-5000, Namur,

Belgium, [email protected]

1 - Beyond the Nash Equilibrium

Bill Corley, Professor, UT-Arlington, Box 19017, Arlington, TX,

76019, United States of America, [email protected]

The Nash, or Regret, Equilibrium remains the standard solution concept for nperson game theory. Two alternative approaches are presented. The

Disappointment Equilibrium (DE) is first defined as a solution where a mutual standoff forces cooperation yielding better outcomes for all players in many situations. The Compromise Equilibrium (CE) is next defined by a scalarization. It gives solutions similar to the DE, but a CE is Pareto optimal as well as computationally tractable for n > 2.

2 - Computing Pure Strategy Equilibria of a Finite n-Person Game

Ping Zhao, PhD Student, City University of Hong Kong, SEEM,

Kowloon, Hong Kong-PRC, [email protected],

Chuangyin Dang

It is well known that determining whether a finite n-person game has a pure strategy equilibrium is an NP-hard problem. To tackle this problem, we propose a convex nonlinear 0-1 integer programming approach. Numerical results show that the approach is promising.

3 - On Traveling Salesman Games with Asymmetric Costs

Nelson Uhan, Assistant Professor, United States Naval Academy,

Mathematics Department, Chauvenet Hall, Annapolis, MD, 21402,

United States of America, [email protected], Alejandro Toriello

We consider cooperative traveling salesman games with non-negative asymmetric costs satisfying the triangle inequality. Using a variant of the Held-Karp relaxation and its dual, we construct a stable cost allocation with budget balance guarantee equal to the Held-Karp integrality gap for the asymmetric traveling salesman problem. Our results extend similar work for traveling salesman games with symmetric costs.

4 - Testing Bidding Equilibrium with Capacity Constrained Auctions

Aadhaar Chaturvedi, Assistant Professor, University of Namur

(FUNDP), Rempart de la Vierge 8, Bureau 406, B-5000, Namur,

Belgium, [email protected], Constantin Blome

In this paper we formulate the equilibrium bidding strategies in a first price sealed bid procurement auction when the suppliers (bidders) are capacity constrained.

We then conduct sensitivity analysis of the equilibrium bid with respect to the capacity of the bidder, the competing supplier and the reserve price announced by the buyer. Finally, we experimentally test our theoretical predictions through controlled laboratory experiments.

TC47

47- North 230- CC

Dynamic Traffic Assignment III - User Equilibrium

Modeling in Continuous-Time

Sponsor: Transportation Science & Logistics/ Intelligent

Transportation Systems (ITS)

Sponsored Session

Chair: Jeff Ban, Assistant Professor, Rensselaer Polytechnic Institute,

110 8th St, JEC 4034, CEE, Troy, NY, 12180, United States of America, [email protected]

1 - Lagrangian-based Hydrodynamic Model: Highway

Traffic Estimation

Tao Yao, Pennsylvania State University, 349 Leonhard Building,

University Park, PA, 16802, United States of America, [email protected], Ke Han, Terry Friesz

The widely studied Lighthill-Whitham-Richards model is formulated in Eulerian coordinate system. We formulate the LWR model in the Lagrangian coordinates and propose a numerical algorithm for solving the PDE with multiple value conditions. This is applied to highway mobile sensing, data fusing and traffic estimation.

329

TC48

2 - Approximating Time Delays in Solving Continuous-time

Dynamic User Equilibrium

Rui Ma, Rensselaer Polytechnic Institute, Troy, NY,

United States of America, [email protected], Jeff Ban, Jong-Shi Pang

An analytic dynamic user equilibrium formulation usually contains time delayed terms to capture realistic traffic dynamics, such as flow propagation. Such delay is often time-varying and state-dependent. In this talk, we propose some schemes to approximate such complicated delayed terms with constant delayed terms which are easier to deal with mathematically. Numerical results are provided to illustrate the proposed approximation method.

3 - Continuous-time Dynamic User Equilibrium Models with

Capacitated Point Queues

Jeff Ban, Assistant Professor, Rensselaer Polytechnic Institute, 110

8th St., JEC 4034, CEE, Troy, NY, 12180, United States of America, [email protected], Rui Ma, Jong-Shi Pang

We propose a differential variational inequality (DVI) formulation for the continuous-time dynamic user equilibrium problem. The formulation integrates the double queue model for capturing more realistic queue dynamics.

4 - Reconstructing the Path-delay Operator for the Continuous

Time Dynamic User Equilibrium Problem

Eric Richardson, Rensselaer Polytechnic Institute, Troy, NY,

United States of America, [email protected], Jeff Ban

In this paper we consider the path-based infinite dimensional variational inequality formulation of dynamic user equilibrium put forward by Friesz et al

1993 and the link-node based dynamic complementarity system formulation by

Ban et al 2011. We show that by properly defining the entry flow function put forth by Friesz et al 1993, we can reformulate the infinite dimensional variational inequality put forward by Friesz et al 1993 into a dynamic complementarity system of the form used by Ban et al 2011. Using this result, we show that the path-delay operator proposed by Friesz et al 1993 can be reconstructed analytically from the link-queue dynamics of Ban et al 2011 providing a continuous time, analytical formulation of the path-delay operator.

TC48

INFORMS Phoenix – 2012

48- North 231 A- CC

Facility Logistics II

Sponsor: Transportation Science & Logistics/ Facility Logistics

Sponsored Session

Chair: Sunderesh Heragu, Professor, University of Louisville,

Department of Industrial Engineering, JB Speed School of Engineering,

Louisville, KY, 40292, United States of America, [email protected]

1 - Optimal Work Release Policies for Order Fulfillment Operations

Kevin Gue, Auburn University, Industrial & Systems Engineering,

Auburn, AL, United States of America, [email protected],

Erdem Ceven

Order fulfillment firms frequently offer service promises of the sort, “Order by this time, and get your shipment the next day.” How do they determine “this time?”

We show how to determine release times for waves in an order fulfillment center to allow “this time” to be as late as possible, thus providing the best service to customers and the greatest competitive advantage to the firm. We illustrate the method with data from a large distributor in the Northeast.

2 - Pareto Optimal Solutions for a Double Row Layout Problem

Chase Murray, Assistant Professor, Industrial & Systems

Engineering, Auburn University, Auburn, AL, United States of

America, [email protected], Xingquan Zuo, Alice E. Smith

We consider a facility layout problem where rectangular machines of unequal size are placed on either side of an aisle. Objectives include minimizing the total material flow cost and minimizing the facility area, with the added restriction that each pair of machines must observe a minimum required clearance. Only problems of very small size may be solved optimally via commercial integer programming solvers. Therefore, a new approach to estimating the Pareto front is proposed.

3 - An Optimal Resource Allocation Model Aging for Bus Transit

Fleet Based on Service Life and Budget Constraint

Sushant Sharma, Purdue University, NEXTRANS, 3000 Kent

Avenue, West Lafayette, IN, 47906, United States of America, [email protected], Sabyasachee Mishra

An optimization model for allocation of funds among different fleet improvement programs within budget constraints over the planning period is presented. This is achieved by minimizing the Net Present Cost of the investment within the constraint of remaining life of the fleet. The model formulation and application are demonstrated with a real world case study of transit agencies.

4 - An Optquest Optimization of a Defense Logistics Warehouse

Sunderesh Heragu, Professor, University of Louisville, Department of Industrial Engineering, JB Speed School of Engineering,

Louisville, KY, 40292, United States of America, [email protected], Banu Ekren

In this study, we propose a near optimum design for the receiving area of the largest warehouse of The Defense Logistics Agency located in New Cumberland,

PA. We develop a simulation based optimization model using OptQuest in ARENA

13.9 software. In the optimization model, we aim to optimize the number of workers required in the induction stations. The performance measure of the system is considered as the average cycle time of a material type having high priority in process.

TC49

49- North 231 B- CC

City Logistics Impact Evaluations

Sponsor: Transportation Science & Logistics

Sponsored Session

Chair: Erica Wygonik, University of Washington, Box 352700/

More Hall, Civil & Environmental Engineering, Seattle, WA, 98195,

United States of America, [email protected]

1 - Economic Impact Analysis of City Logistics Alternatives using the Southern California Planning Mode

Qian An, PhD Candidate, University of Southern California,

Los Angeles, CA, 90007, United States of America, [email protected],

James Moore, Maged Dessouky

Two-level city logistics distribution system has the potential to reduce distribution cost, but the most importantly benefit is the improvement of traffic conditions and other aspects of urban life impacted by the indirect cost of vehicles, such as environmental quality and safety. We apply the Southern California Planning

Model, which is a spatial input-output model of the urban economy, to measure regional economic impacts in, but not limited to, the five-county Los Angeles region.

2 - How Carbon Pricing Affects the Efficiency of Urban Delivery

Cooperation?

Qin Chen, University of Illinois-Chicago, 847 W Taylor St.,

MC246, Chicago, IL, 60607, United States of America, [email protected], Jane Lin

We analyze the impact of carbon pricing on urban delivery cooperation by comparing the cost of delivery with/without cooperation. A combination of logistics cost and carbon emission cost is considered in the objective. MOVES model is used to capture the transportation emissions in the logistics activities.

The delivery problem is formulated at the tactical level with Continuous

Approximation method, which approximates the discrete data with continuous function in a closed form solution.

3 - Commercial Vehicles Fleet Replacement Optimization:

Electric Trucks vs Conventional Diesel Trucks

Miguel Figliozzi, Associate Professor, Portland State University,

P.O. Box 751 CEE, OR 97207-075, Portland, OR, United States of

America, [email protected], Wei Feng

This research presents a vehicle fleet replacement model that aims at minimizing fleet total net cost in a planning time horizon, including purchase cost, energy cost, operating and maintenance cost, emissions cost and salvage revenue. A set of economic and technological factors, as well as fleet initial configuration are incorporated into the model, and their impacts on the competitiveness between electric trucks and conventional trucks are analyzed.

4 - Developing Heuristics for the VRP to Address Emissions, Time

Windows, and Fleet Heterogeneity

Erica Wygonik, University of Washington, Box 352700/

More Hall, Civil & Environmental Engineering, Seattle, WA,

98195, United States of America, [email protected],

Felipe Sandoval, Anne Goodchild

This work extends existing heuristics to accommodate the additional complexity of emissions, time windows, and fleet heterogeneity on solutions to the Vehicle

Routing Problem. The model is tested and refined using real-world case studies to evaluate the influence of the differing constraints on the solutions. Each case study is unique with differing vehicle heterogeneity, network connectedness, and load characteristics.

330

TC50

50- North 231 C- CC

Modeling and Analyzing Military Operations and

Systems II

Sponsor: Military Applications

Sponsored Session

Chair: Amnon Gonen, Dean, Faculty of Management of Technology,

HIT - Holon Institute oof Technology, 52 Golomb St., Holon, 58102,

Israel

1 - Military Reality Modeling and Approximation

Amnon Gonen, Holon Institute of Technology, HIT, Golomb 52,

Holon, Israel, [email protected]

Some of the military models are so complicated that users can hardly estimate the magnitude of results. The paper discusses approximation techniques to hit and kill probably. In some cases the physical model of human behavior is approximated by an “inverse engineering” approach that will be shown by the suppression modeling. In this case, the modeling starts from the results toward their cause and formulate the calculation process that fits these final results.

2 - Mathematical Modeling for Decision Making

William Fox, Naval Postgraduate School, Monterey, CA, 93943,

United States of America, [email protected]

Building models to inform decision makers about alternative choices is critical in today’s rapidly changing world. In this presentation we examine several scenarios and illustarte how mathematical modeling can be used as a decision tool to assist the decision maker.

3 - Soldier Narrative Analysis as Part of a Rapid Fielding Process

Michael Jaye, Associate Professor, Department of Defense Analysis,

1 University Way, Monterey, CA, 93943, United States of America, [email protected], Patrick Driscoll

The warfighter Technology Tradespace Methodolgy employs technology acceptance theories and applies them to TSOA activities, which focus on working with developers to gather system data relative to warfighter expectations and concerns. Collecting and analyzing narratives from soldier end users can inform that process. The accumulation and tagging of multiple sense-making items provide sophisticated metadata which can be used to provide quantitative research material, as well as impact analysis.

TC51

INFORMS Phoenix – 2012

51- North 232 A- CC

Big Data Analytics for Condition Based Maintenance

Sponsor: Military Applications

Sponsored Session

Chair: Dan Widdis, Fellow, Principal Operations Research Analyst,

Concurrent Technologies Corporation, 5897 Castleberry Peak Ave.,

Las Vegas, NV, 89131, United States of America, [email protected]

1 - Using Large Data Analytics and Prognostics Simulation for

Performance-Based Investment Decisions

Norm Reitter, Advisor, Information Technology, Concurrent

Technologies Corporation, 329, 44th Street, Pittsburgh, PA, 15201,

United States of America, [email protected]

Investments in Army CBM aviation health monitoring devices to collect sensor data provide an opportunity to reduce supply chain costs through improved prognostic capabilities. Gleaning useful prediction capability from these very large data sets requires robust experimentation and simulation to sufficiently test the benefits of these investments. We use predictive algorithms to determine a combined risk of stock out and simulation methods to determine expected supply chain performance.

2 - A Big Data CBM Solution using Embedded Analytics

Natarajan Sridhar, Advisor Information Architect, CTC,

100 CTC Drive, Johnstown, PA, 15963, United States of America, [email protected]

There are multiple challenges associated with implementing embedded analytics in big data CBM environments to support decisions across multiple maintenance/supply echelons (Unit to Enterprise):(a) managing “big” source data;(b) provisioning focused analytic datasets; and(c) providing timely/effective decision support information. We present a solution that combines systems engineering and analytics best practices to support unit-level decision support for

Army Aviation platforms.

3 - Semantic Reasoning Over Big Data for Condition Based

Maintenance and Mission Based Forecasting

Wes Regian, Principal Knowledge Engineer, Concurrent

Technologies Corporation, 15091 Alabama Highway 20, Suite A,

Madison, AL, 35756, United States of America, [email protected]

CBM as practiced today ignores impending mission characteristics. Brigade

Aviation Maintenance Officers (BAMOs) on the other hand, predict parts needs for upcoming mission characteristics based on failures observed in the past under similar mission conditions. Semantic analysis offers the possibility of machine reasoning to combine knowledge of MBF with CBM – supporting BAMO-like machine reasoning in combination with CBM capabilities.

TC52

52- North 232 B- CC

Decision Making for Electric Vehicles

Sponsor: Energy, Natural Res & the Environment/Energy

Sponsored Session

TC52

Chair: Timothy Sweda, PhD Candidate, Northwestern University,

2145 Sheridan Rd., Rm. C210, Evanston, IL, 60208, United States of

America, [email protected]

1 - A Robust Dynamic Algorithm to Provide Grid Services with a

Fleet of Plug-In Electric Vehicles

Nicole Taheri, Stanford University, 344 Olmsted Road, #449,

Stanford, CA, 94305, United States of America, [email protected], Robert Entriken, Yinyu Ye

Plug-in electric vehicles (PEVs) may be capable of discharging and flexible charging options. We construct an algorithm for a fleet of PEVs that efficiently organizes distributed electricity trading to benefit consumers and utilities. A robust linear programming model of the fleet can be used by an aggregator to allocate energy exchange schedules to vehicles instantly as they plug-in. The resulting schedules are robust to uncertain driving behavior. We give empirical results using real data.

2 - Approximate Dynamic Programming for Battery Purchase and

Charging at Swap Stations

Frank Schneider, PhD Candidate, University of Cologne, Albertus-

Magnus-Platz, Department for SCM & Mgmt. Science, Cologne,

50923, Germany, [email protected], Diego Klabjan

We formulate and solve the decision problem of the number of batteries to have in circulation at a battery swapping stations. This also entails the decision making about how many batteries to start charging in each time period. Demand for battery swapping and prices of electricity are stochastic. We propose an algorithm combining classical optimization and approximate dynamic programming. We also provide a numerical study.

3 - Optimal Routes for Electric Vehicles Facing Uncertainty,

Congestion, and Energy Constraints

Matthew Fontana, Massachusetts Institute of Technology, 77

Massachusetts Avenue, Cambridge, MA, United States of America, [email protected], Daniel Reich, Oleg Gusikhin, Perry MacNeille,

Dimitris Bertsimas, Erica Klampfl, Thomas Magnanti, Ryan McGee

To address range anxiety for drivers of battery electric vehicles, we propose a robust optimization model to find optimal routes under uncertainty and congestion while considering constraints on total energy. We describe the physics of electric vehicles, the overall model, and our solution approach. We have implemented our approach using data from a Geographic Information System with very promising results.

4 - Optimal Routing and Recharging Algorithms for Electric

Vehicles

Timothy Sweda, PhD Candidate, Northwestern University,

2145 Sheridan Rd., Rm. C210, Evanston, IL, 60208, United States of America, [email protected], Diego Klabjan,

Irina Dolinskaya

Recharging decisions for electric vehicles (EVs) are complex for a number of reasons. The time required to recharge can be significant compared to the total trip time, and the charge rate at a given charging station depends on the station’s voltage and the vehicle’s charge level. In addition, battery health must be preserved. We model the problem of coordinating routing and recharging decisions for EVs as a dynamic program and present several solution algorithms and their analyses.

331

TC53

TC53

53- North 232 C- CC

Pricing in Electricity Markets Including

Nonconvexities and Stochastic Producers

Sponsor: Energy, Natural Res & the Environment/Energy

Sponsored Session

Chair: Antonio Conejo, Professor, Universidad Castilla - La Mancha,

Campus Universitario, Ciudad Real, 13071, Spain,

[email protected]

1 - A Single Settlement Energy only Market to Harness

Wind Generation

Golbon Zakeri, University of Auckland, 70 Symonds Street,

Auckland, New Zealand, [email protected], Geoff Pritchard,

Andy Philpott

We discuss a stochastic-programming-based method for scheduling electric power generation subject to uncertainty. Such uncertainty may arise from either imperfect forecasting or moment-to-moment fluctuations, and on either the supply or the demand side. The method gives a system of locational marginal prices which reflect the uncertainty, and these may be used in a market settlement scheme in which payment is for energy only. We show that this scheme is revenue-adequate in expectation.

2 - Pricing Electricity in Pools with Stochastic Producers

Juan Miguel Morales, Technical University of Denmark, Kongens

Lyngby, Denmark, [email protected], Antonio Conejo

As the presence of stochastic producers in a power system increases, balancing costs may become highly dependent on the day-ahead market outcomes. We present a clearing algorithm for a day-ahead electricity market that co-optimizes the forward dispatch and the projected balancing actions by using a two-stage stochastic programming approach. A set of electricity prices that consistently supports the forward positions acquired by market agents under this form of market clearing is investigated.

3 - Pricing Non-convexities in an Electricity Pool

Carlos Ruiz, École Centrale Paris and Supelec,

Grande Voie des Vignes, Châtenay-Malabry, 92295, France, [email protected], Antonio Conejo, Steven Gabriel

Electricity pools are generally cleared through auctions formulated as non-convex

MILP problems that make difficult the derivation of adequate marginal prices. In this work we propose a primal-dual approach to derive efficient revenue adequate uniform prices that guarantee that dispatched producers are willing to remain in the market. Such prices may not significantly deviate from the marginal prices obtained if integrality conditions are relaxed in the original MILP problem.

TC54

54- Regency Ballroom A- Hyatt

Learning in Social Networks

Sponsor: Applied Probability

Sponsored Session

INFORMS Phoenix – 2012

Chair: Ilan Lobel, New York University, Stern Business, 44 W 4th St,

New York, NY, United States of America, [email protected]

1 - Observational Learning with Finite Memory

Kimon Drakopoulos, Massachusetts Institute of Technology, 32

Vassar St., Cambridge, MA, 02139, United States of America, [email protected], John N. Tsitsiklis, Asuman Ozdaglar

We consider a sequence of agents making decisions regarding the unknown value of the underlying state of the world after receiving private information and observing the decisions of their K immediate predecessors. We study conditions on the communication and information structure under which agents’ decisions converge to the correct value of the underlying state. Finally , we analyze the learning behavior of the equilibria that arise from a sequential game between forward looking agents.

2 - Bayesian Social Learning with Consumer Reviews

Bar Ifrach, Columbia Business School, 3022 Broadway, Uris Hall, 4

West, New York, NY, 10027, United States of America,

[email protected], Marco Scarsini, Costis Maglaras

We study a market with heterogeneous customers who rationally learn the quality of an offered product by observing the reviews of customers who purchased the product earlier in time. We explore aspects of this social learning process such as: its success in revealing the true quality over time, and the effect of the order in which reviews are submitted on consumers’ beliefs. In addition, we draw conclusions on the seller’s pricing problem when accounting for social learning.

3 - Social Learning with Network Uncertainty

Evan Sadler, New York University, Stern Business, New York, NY,

United States of America, [email protected], Ilan Lobel

We construct a sequential model of social learning in complex networks and examine the perfect Bayesian equilibria of this model. In contrast to prior models in which the neighborhoods of agents are independent of each other, we consider arbitrary network topologies and show that prior characterizations of conditions under which learning occurs fail. We give new characterizations that offer insights on new phenomena that can occur in complex networks.

4 - A Bayesian Approach for Predicting the Popularity of Tweets

Tauhid Zaman, Massachusetts Institute of Technology, 77

Massachusetts Ave., Cambridge, MA, United States of America, [email protected], Emily Fox, Eric Bradlow

We present a Bayesian approach for predicting the popularity of “tweets” from the micro-blogging site Twitter. Our approach predicts how many people will interact with a tweet, along with how quickly its popularity will rise using only observations of when users interact with the tweet and their basic social network information. We find our method can produce relatively accurate predictions even when observing only a small number of interactions.

TC55

55- Regency Ballroom B - Hyatt

Undergraduate Operations Research Prize II

Cluster: Undergraduate Operations Research Prize

Invited Session

Chair: Feryal Erhun, Stanford University, Stanford, CA,

United States of America, [email protected]

1 - A Cost-minimizing Heuristic Solution Algorithm for the Network

Sensor Location Problem

Emanuel Seitinger, Vienna University of Economics and Business,

Vienna, Austria, [email protected]

The cost-effective and accurate estimation of traffic flows is of immense importance in traffic planning. We discuss one exact and one heuristic solution procedure for the network sensor location problem. While the exact solution procedure turns out to be unfeasible for many real-life applications, its modification that takes costs, assumptions on human behavior and turning proportions into account comes up with greatly improved results.

2 - A Heuristics for Radio Station Location for Telecommunications in Rural Areas

José Ignacio del Villar, Instituto Tecnológico Autónomo de México,

Mexico City, Mexico, David Fernando Muñoz

We discuss the problem of finding a minimum cost location of Long-Haul equipment for wireless telecommunications in rural areas. We show that this problem has an Integer Linear Programming formulation. We developed and tested a heuristics based in the ideas of the Minimum Spanning Tree. Our experiments show that the solution provided by this heuristics may be used as a good initial solution when developing an exact algorithm or a search algorithm specially designed for this problem.

3 - Network Theoretic Characterization of Corporate Structures

Hao Fu, Vassar College, 124 Raymond Avenue, Box 2555,

Poughkeepsie, NY, 12604, United States of America, [email protected]

It is unclear whether traditionally hierarchal corporate structure meets today’s demand for innovation. This research provides a framework for comparing different corporate structures using network theoretic characterizations. Some prevalent social networks are examined and compared to hierarchical structure both analytically and numerically. The result shows structures resembling social networks had more advantage in innovation. A general algorithm is provided to design desirable structures.

4 - An Access Network Design Problem With Nonlinear Qos

Constraints

Chanwoo Park, Korea University, Anamdong 5 Ga, Sungbuk Gu,

Seoul, Korea, Republic of, [email protected]

We present an access network design problem with nonlinear QoS constraints.

The problem can be conceptualized as a two-level hierarchical location-allocation problem on the tree topology. By exploiting the special structure of the nonlinear

QoS constraints, we devise an exact optimal algorithm with an effective constraint generation strategy. We present promising computational results that demonstrate the effectiveness of the proposed solution procedure.

332

INFORMS Phoenix – 2012

TC56

56- Curtis A- Hyatt

Emerging Themes in Technology Management

Sponsor: Technology Management

Sponsored Session

Chair: Cheryl Druehl, George Mason University, Fairfax, VA,

United States of America, [email protected]

1 - A Model of Product Specification Decision-making and

Project Failure

Zhijian Cui, Assistant Professor of Operations Management, IE

Business School, Calle de Maria de Molina, 12, Madrid, 28006,

Spain, [email protected], Christoph Loch

This study models the bargaining process between two parties (leader and follower) on a one-dimensional feature design decision for a new product. We examine three managerial levers to reduce the risk of bargaining failure. First, follwer can reduce the failure risk by signaling his fallback. Second, giving leader an incentive (as with a bonus) exacerbates the risk of failure. Third, if follower cares about his relative payooff or status, this also increases the risk of failure.

2 - New Wine in Old Bottles? Print-newspapers’ Overlooked Role in

Digital Media Diffusion

Bo Kyung Kim, Assistant Professor, Southern Methodist

University, Cox School of Business, 6212 Bishop Boulevard 323

Fincher, Dallas, TX, 75205-0006, United States of America, [email protected]

During the implementation process of discontinuous technology, incumbents can emphasize the similarity between new and current technologies to have better adoption performance by making their current assets more valuable in the new market. In the U.S. daily newspaper industry, incumbent newspapers that deemphasized interactivity on the web, the most distinctive character of digital media from print media, indeed had better online readership, potentially creating

“print-media-like” digital media.

3 - How Can Companies Leverage Their External Environment when Engaging in Open Innovation?

Irina Savitskaya, Post-Doctoral Researcher, Lappeenranta

University of Technology, Skinnarilankatu 34, Lappeenranta,

53850, Finland, [email protected]

In the modern environment firms do not operate by themselves, they are rather a part of ecosystems. Hence their actions and success depends on how good are they in co-innovation. When companies engage in open innovation activities, they most commonly build a strategy on cooperation with the direct stakeholders.

What often stays neglected is the role of external factors influencing and shaping this cooperation: structural, institutional and cultural environment play an important role.

4 - Paradigm-changing vs. Paradigm-deepening Innovation:

Firm Scope and Technological Response to Shocks

Vivek Tandon, Assistant Professor, National University of

Singapore, 15 Kent Ridge Drive, Singapore, Singapore, [email protected], Gautam Ahuja, Curba Lampert

We examine firms’ technological response to supply shocks (sharp increase in a key input’s price). Firms can respond by investing in paradigm-changing technologies that use substitute inputs or in paradigm-deepening ones that improves the efficient use of existing inputs. Our framework relates this choice with the degree of relatedness across a firm’s businesses. We test our hypotheses examining the responses of large manufacturing firms in the United States to the oil shock of the early 1980s.

TC57

57- Curtis B- Hyatt

Applied Probability in Energy and Risk Analysis

Sponsor: Applied Probability

Sponsored Session

Chair: Jose Blanchet, Columbia University, 500W 120th St, New York,

NY, United States of America, [email protected]

1 - Searching for Optimal Control Algorithms for a Power Grid

System using Efficient Simulation Methods

Juan Li, Columbia University, IEOR Department, 500 West 120th

Street, New York, NY, 10027, United States of America, [email protected], Jose Blanchet, Daniel Bienstock

Considering a power grid system with random wind power and stochastic line outages, we study the optimal control in the rare event of a cascading failure using efficient simulation methods. We implement our optimal control algorithms into real-world examples. Numerical results are provided.

TC58

2 - Networking Low-power Energy Harvesting Devices:

Measurements and Algorithms

Aya Wallwater, Columbia University, Rm. 325 Mudd Building,

500 W. 120 Street, New York, NY, 10027, United States of America, [email protected], Gil Zussman, Maria Gorlatova

Recent advances in energy harvesting will soon enable networks composed of energy harvesting devices.These devices will operate using very low energy.We

characterize the light energy available in indoor environments and present new algorithms, which requires non-traditional approaches, since energy harvesting shifts the nature of energy-aware protocols from minimizing energy expenditure to optimizing it. We develop algorithms for systems with predictable energy inputs, as well as stochastic.

3 - A Stochastic Model for Risk Management

Xinyun Chen, Columbia University, 500 West 120th Street, 323

Mudd Building, New York, NY, 10027, United States of America, [email protected], Jose Blanchet

We develop a bottom-up model for risk analysis of a family of financial systems.

The model is able to capture the changes in the macro-environment and the dependence among individuals in the system. We also show some convergence results of the related stochastic processes and as a direct consequence, many quantities of interest, such as the default probability, can be estimated from the model parameters.

TC58

58- Phoenix East- Hyatt

Healthcare Operations

Sponsor: Applied Probability

Sponsored Session

Chair: Nikos Trichakis, Assistant Professor, Harvard Business School,

Boston, MA, United States of America, [email protected]

Co-Chair: Ross Anderson, Massachusetts Institute of Technology,

77 Massachusetts Avenue, Cambridge, MA, 02142, United States of

America, [email protected]

1 - Efficiency of Joint-ventures in the Healthcare Industry

Retsef Levi, Massachusetts Institute of Technology, 100 Main

Street, Building E62-562, Cambridge, MA, United States of

America, [email protected], Georgia Perakis, Cong Shi, Wei Sun

Motivated by the growing popularity of joint ventures among healthcare providers, we consider a setting with demand uncertainty and compare the joint venture profit with nonlinear convex costs in a Nash equilibrium to a system optimum which maximizes the collective benefit. We develop a parametric bound on the efficiency of joint ventures. We show that the efficiency depends on the marginal cost to revenue ratio as well as the spread of the demand distribution.

2 - A Dynamic Model of Kidney Exchange Programs

Ross Anderson, Massachusetts Institute of Technology, 77

Massachusetts Avenue, Cambridge, MA, 02142, United States of

America, [email protected], Itai Ashlagi, David Gamarnik

Kidney exchanges enable incompatible donor-patient pairs to trade donors with other incompatible pairs. In practice, kidney exchanges periodically compute a set of trades to maximize the number of transplants. However, these policies do not consider long term objectives. For a dynamic model, we show that if trades are assigned too frequently, then the average waiting time for a transplant actually increases. We corroborate this result with historical data from a large kidney exchange.

3 - Multi-hospital Kidney Exchange: Dynamic Matching in a

Heterogenous System

Vahideh Manshadi, Postdoc, Massachusetts Institute of Technology,

77 Massachusetts Avenue, Bldg. E40-113, Cambridge, MA, 02139,

United States of America, [email protected], Patrick Jaillet,

Itai Ashlagi

Matching algorithms are major tools for kidney exchange programs and different programs use different algorithms. A major unresolved concern is the tradeoff between waiting time and number of matches. In this work we begin to develop mechanisms for kidney exchange in large kidney exchange programs in a dynamic setting. Our goal is to devise allocation rules for which the average waiting time of pairs is small and at the same time, the total number of exchanges is large.

333

TC59

INFORMS Phoenix – 2012

TC59

59- Phoenix West- Hyatt

Optimal Control Policies II

Sponsor: Applied Probability

Sponsored Session

Chair: Ananth Krishnamurthy, University of Wisconsin-Madison, 1513

University Avenue, ME 3258, Madison, WI, United States of America, [email protected]

1 - Optimal Control of an Intersection

Erjen Lefeber, Technische Universiteit Eindhoven, Eindhoven,

5600MB, Netherlands, [email protected], Dirk van Zwieten

In this paper we present a method to determine an optimal periodic orbit for a fluid model describing an intersection of urban traffic. We first determine the set of feasible modes (clique problem), subsequently the possible sets of modes to use during a cycle, then the sequence of modes (traveling salesman problem), and finally the duration of each mode (quadratic program). We apply this method to a crowded intersection in the city of Eindhoven.

2 - Optimal Harvesting Policies to Control Batch-to-batch

Variability in Bioreactors

Tugce Martagan, University of Wisconsin-Madison,

1550 Engineering Drive, Engineering Cent, Madison, WI, 53706,

United States of America, [email protected]

Controlling batch-to-batch variability presents a significant challenge in biomanufacturing processes. In this paper, we model upstream biomanufacturing operations using Markov Decision Process model, and identify optimum harvesting policies to minimize expected cost in the presence of batch-to-batch variability. We subsequently assess the robustness of these policies to manufacturing uncertainties.

3 - Optimal Control of a Two-server Queueing System with Breakdowns

Jeffrey Kharoufeh, Associate Professor, University of Pittsburgh,

Department of Industrial Engineering, 1048 Benedum Hall, 3700

O’Hara Street, Pittsburgh, PA, 15261, United States of America, [email protected], Erhun Ozkan

We consider controlling a two-server Markovian queueing system with heterogeneous servers, at least one of which is unreliable. Our objective is to dynamically route customers at arrival, failure, and repair epochs to minimize the long-run average number of customers in the system. We show the existence of an optimal threshold-type policy that depends on the number in queue and the status of the servers.

TC60

60- Remington- Hyatt

Impacts of Environmental and Quality Programs in the Airline Industry

Sponsor: Aviation Applications

Sponsored Session

Chair: Kushal A. Moolchandani, Purdue University, School of

Aeronautics and Astronautics, 701 W. Stadium Avenue, West Lafayette,

IN, United States of America, [email protected]

1 - Assessing the U.S. Aviation Environmental Impacts using a

Duopoly Airlines Model

Kushal A. Moolchandani, Purdue University, School of

Aeronautics and Astronautics, 701 W. Stadium Avenue, West

Lafayette, IN, United States of America, [email protected],

Satadru Roy, Datu B. Agusdinata, Muharrem Mane, Daniel

DeLaurentis, William Crossley

The operations of legacy and low-cost carriers in the U.S are simulated under various scenarios of economic conditions and policy implementations. The duopoly model is solved as a resource allocation set up to maximize airlines’ profit while meeting market demand. This work also incorporates the ability to model airlines’ decisions on retirement and acquisition of aircraft. This study helps assess the resulting emissions and noise impacts and could inform technological development and formulation of policies intended to reduce them.

2 - Impacts of Deploying Six Sigma Quality Control in Airline

Operations

Ramesh Bollapragada, Professor, College of Business, San

Francisco State University, 1600 Holloway Avenue, San Francisco,

CA, 94132, United States of America, [email protected],

Vivian Chan

The purpose of the paper is to investigate the financial and operational benefits when deploying Six Sigma (SS) methodologies in airlines operations. The Six

Sigma continuous improvement methodology DMAIC is used in this study. SS statistical calculations are used to measure the current performance of each critical metric involving operations processes within international and U.S.

domestic airlines industry.

3 - Analysis of Economic and Environmental Impact of New Aircraft

Concepts via Allocation and Scheduling

Isaac Tetzloff, Graduate Student, Purdue University, 701 W.

Stadium Ave., West Lafayette, IN, 47907, United States of America, [email protected], William Crossley

Engineering analyses can predict emissions of new aircraft or technologies; however, evaluating their impact on airline profit and the environment at the fleet-level requires studying the interactions between the new aircraft and the airline’s current fleet. Through the use of an aircraft allocation and scheduling tool, decisions on how to deploy new aircraft technologies and concepts within an airline’s fleet are quickly made while considering both airline profit and environmental impact.

4 - Does EU ETS Instigate AirCargo Network Reconfiguration?

A Model-based Analysis

Ulrich Derigs, Department of Information Systems and Operations

Research (WINFORS), University of Cologne, Cologne, Germany, [email protected], Stefan Illing

From 2012 on aviation is included in the European Emissions Trading Scheme

(EU ETS) and operators have to hold one allowance per tonne of CO2 emitted on every flight departing from and/or arriving at an airport within the EU. Now two questions are of interest: Is it profitable for airlines to reconfigure their routes to reduce EU-related emissions and costs, and, will the schema be successful in the sense that emissions are reduced significantly. Here the potential for and the consequences of reconfiguration are different for the passenger and cargo business, respectively. In this paper we present a model-based simulation of network (re-)configuration/optimization at cargo airlines under different EU ETS scenarios and we discuss the results with respect to the two issues raised above.

TC61

61- Russell- Hyatt

Patrolling Problems in Security

Cluster: Applications in Emergency Management and Terrorism

Security

Invited Session

Chair: David Alderson, Operations Research Department,

Naval Postgraduate School, Monterey, CA, United States of America, [email protected]

1 - Game Theory for Security: Key Algorithmic Principles, Deployed

Applications, Lessons Learned

Milind Tambe, University of Southern California, Los Angeles, CA,

United States of America, [email protected]

We will provide an overview of key principles of several game-theoretic applications developed by my research team at USC that are now deployed over multiple years with multiple security agencies: LAX police, Federal Air Marshals

(FAMS), US coast guard in Boston and New York, and Los Angeles Sheriff

Department.

2 - Security Games with Limited Surveillance

Bo An, University of Southern California, Los Angeles, CA,

United States of America, [email protected], Yevgeniy Vorobeychik,

David Kempe, Christopher Kiekintveld, Milind Tambe, Eric Shieh,

Satinder Singh

We model the process of an attacker observing a sequence of resource allocation decisions and updating his beliefs about the defender’s strategy. We present exact and approximate computational techniques for updating the attacker’s beliefs and computing optimal strategies for both the attacker and defender, given a specific number of observations.

334

3 - A Patrol Problem from the Adversary’s Perspective

Kyle Lin, Associate Professor, Naval Postgraduate School, 1411

Cunningham Rd, Monterey, CA, 93943, United States of America, [email protected], David Alderson, Michael Atkinson, Tim Chung

Consider a patrol problem, where a defender uses patrols to deter attacks, while an adversary observes and studies the patrol strategy before deciding whether to attack. We use a game-theoretic model to study these interactions and draw insights into the effectiveness of patrols.

TC62

62- Borein A- Hyatt

Auctions and the Core

Cluster: Auctions

Invited Session

Chair: Robert Day, Associate Professor, OPIM, University of

Connecticut, 2100 Hillside Road, Unit 1041, Storrs, CT, 06269,

United States of America, [email protected]

1 - Risk Aversion in Ascending Core-selecting Auctions

Ioannis Petrakis, TU München, Room 01.10.055, Boltzmannstr 3,

Munich, 85748, Germany, [email protected], Kemal

Guler, Martin Bichler

We characterize conditions for the perfect Bayesian equilibria of the ascending core-selecting auction mechanism to have the small bidders to drop at the reserve price with risk-averse bidders. Our first main result is a generalization of the condition for a non-bidding equilibrium in (Sano, 2011), which allows for arbitrary concave utility functions. Second, we discuss this condition in the presence of asymmetries, and third, we provide comparative statics.

2 - Core Deviation Minimizing Auctions

Isa Hafalir, Carnegie Mellon University, Tepper School of Business,

5000 Forbes Ave., Pittsburgh, PA, 15217, United States of America, [email protected]

We study dominant strategy implementable direct mechanisms that minimize the expected surplus from core deviations. Using incentive compatibility conditions, we formulate the core deviation miminimization problem as a calculus of variations problem and then numerically solve it for some particular cases.

3 - Core-selecting Auctions with Incomplete Information

Oleg Baranov, Assistant Professor, University of Colorado at

Boulder, Department of Economics, 256 UCB, Boulder, CO, 80309,

United States of America, [email protected],

Lawrence Ausubel

In this paper, we consider a simple incomplete-information model which allows us to do a full equilibrium analysis for four different core-selecting auction formats suggested in the literature. We find that the revenues and efficiency from core-selecting auctions improve as correlations among bidders’ values increase, while the revenues from the Vickrey auction worsen. Thus, there may be good reasons for policymakers to utilize core-selecting auctions rather than a VCG in realistic environments.

TC63

63- Borein B- Hyatt

Joint Session Behavioral Oper/Auctions:

Behavioral Mechanism Design

Sponsor: Behavioral Operations & Auctions

Sponsored Session

INFORMS Phoenix – 2012

Chair: Elena Katok, Professor, University of Texas Dallas, 800 West

Campbell Rd., Richardson, TX, 75080, United States of America, [email protected]

1 - Supply Chain Contract Design: Impact of Bounded Rationality and Individual Heterogeneity

Diana Wu, University of Kansas, Lawrence, KS,

United States of America, [email protected], Kay-Yut Chen

We model two forms of bounded rationality in a newsvendor setting: the tendency to err and the tendency to anchor. The model is motivated by observations under several types of supply chain contracts in the laboratory. Our experiments show that the behavior of newsvendors follows some multi-modal distribution that responds to contract settings. Moreover, we discover heterogeneity in decision makers’ tendencies. New insights on contract design are obtained by applying this behavioral model.

2 - Commitment and Reason in Bargaining with Joint Production

Gary Bolton, Schwartz Professor of Business Economics,

Pennsylvania State University, 334 Business Building, University

Park, PA, 16802, United States of America, [email protected].psu.edu

We conduct a bargaining experiment with joint production to investigate the relationship between hard leverage (the option to make binding commitments) and soft leverage (appeals to asymmetric focal points) in providing bargaining advantage. Our data is consistent with the hypothesis that reason in the form of focal points bounds the use of commitment.

3 - Investigating Strategic Customer Behavior through an

Interactive Supply Chain Game

Pelin Pekgun, Assistant Professor, University of South Carolina,

Moore School of Business, 1705 College Street, Columbia, SC,

29208, United States of America, [email protected],

Pinar Keskinocak, Tosanwunmi Maku, Mani Janakiram

We investigate the strategic forecasting and ordering behavior of customers through an interactive supply chain game with the motivation of continuously improving supply chain metrics and planning for various scenarios. We discuss the effectiveness of different inventory allocation mechanisms in reducing forecast inflation based on our insights gained from experiments conducted on over 200 human subjects.

4 - Designing Incentive Schemes for Truthful Forecast

Information Sharing

Ulrich Thonemann, University of Cologne, Albertus-Magnus-Platz,

Cologne, 50923, Germany, [email protected],

Lisa Scheele, Marko Slikker

Motivated by structurally biased sales forecasts in practice, we study a forecast sharing situation between a (better informed) sales and a production planning department. Our objective is to incentivize truthful forecasts by applying incentive schemes which penalize forecast errors. We find that experimental observations are at odds with game-theoretic predictions under rationality assumptions and develop a new behavioral model based on theories of reference point dependence and lying aversion.

TC66

TC66

66- Ellis West- Hyatt

Joint Session DM/QSR: Novel Classification and

Clustering Algorithms for Data Mining

Sponsor: Data Mining & Quality, Statistics and Reliability

Sponsored Session

Chair: Seoung Bum Kim, Associate Professior, Korea University,

Seoul, Korea, Republic of, [email protected]

Co-Chair: Jun-Geol BaekAssociate Professor, Korea University, Seoul,

Korea, Republic of, [email protected]

1 - A Recursive Binary Partitioning Algorithm for Clustering

Analysis

Jihoon Kang, PhD Student, Korea University, Seoul, Korea,

Republic of, [email protected], Seoung Bum Kim

Clustering analysis has been conducted to elicit nature groupings of the dataset without prior information about the sample class. In this study, we propose a new clustering algorithm that adopts recursive binary partitioning to achieve the best grouping of the dataset. Experiments with the simulation and real data demonstrate the effectiveness and interpretability of the proposed algorithm.

2 - A Novel Support Vector Data Description Algorithm

Poovich Phaladiganon, PhD Candidate, University of Texas at

Arlington, Arlington, TX, United States of America, [email protected], Seoung Bum Kim,

Victoria Chen

Support vector data description (SVDD) is a one-class classification technique, which generates spherical boundary. The goal is to maximize the number of observations and minimize the radius of sphere. The boundary is determined by support vectors (SVs). However, the boundary generated from SVDD does not consider the data density. Thus, we utilize SVs to incorporate data density into

SVDD. The preliminary research shows that the proposed method performs better than the existing one.

3 - An Ensemble Classification Algorithm with Clustered Features

Chanhee Park, Korea University, Seoul, Korea,

Republic of, [email protected], Seoung Bum Kim

In this study, we propose a new classification algorithm based upon a set of features clustered. The features were clustered by a standard clustering algorithm followed by the classification analysis in each clustered feature. We used the simulation and real data to demonstrate the efficiency and usefulness of the proposed algorithm.

335

TC67

INFORMS Phoenix – 2012

4 - The Analysis of Dengue Virus Serotypes via UnitX

Representation

Chivalai Temiyasathit, Assistant Professor, King Mongkut’s

Institute of Technology Ladkrabang, International College, KMITL,

Chalongkrung Road, Ladkrabang District, Bangkok, 10520,

Thailand, [email protected], Yodchanan Wongsawat

One of the challenging problems in virus sequence analyses is the graphical representation of the long sequences nucleotides. In this study, a new graphical representation based on the cumulative amount of amino and keto bases, called

UnitX method is proposed. Classification tree is then employed to verify the distinguish ability of the projected sequences. Experimental results showed that the proposed method can efficiently distinguish the four serotypes of Dengue virus sequences.

TC67

67- Ellis East- Hyatt

Fault Management for Complex Engineering Systems

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Shiyu Zhou, Professor, University of Wisconsin, Madison,

Department of Industrial Engineering, 1513 University Avenue,

Madison, WI, 53706, United States of America, [email protected]

Co-Chair: Qiang Zhou, Assistant Professor, City University of Hong

Kong, Hong Kong, Hong Kong - PRC, [email protected]

1 - A Framework for Variation Visualization and Understanding in

Complex Manufacturing Systems

Fadel Megahed, Assistant Professor, Auburn University,

3301L Shelby Center, Auburn, Al, United States of America, [email protected], Jaime Camelio, William Woodall, Lee Wells

We provide a framework that allows industrial practitioners to visualize the most significant variation patterns. In essence, this framework complements Phase I techniques by enabling users to: (1) acquire detailed understanding of commoncause variability; (2) quickly visualize the effects of common-cause variability in a process with respect to the final product; and (3) identify opportunities for process improvement. The framework is illustrated through a case study.

2 - Optimal Maintenance of Multi-component Systems Considering

Component Failure Dependency

Qingyu William Yang, Assistant Professor, Wayne State University,

4815 Fourth Street, Rm 2167, Detroit, MI, 48335,

United States of America, [email protected], Nailong Zhang

The interactions between components in multi-component systems complicate the modeling and optimization of maintenance. In this research, we develop optimal maintenance planning for multi-component systems that can capture component failure dependency. A case study of engine assembly system is conducted to verify the developed method.

3 - Quantifying Boundary Effect of Nanoparticles in Metal Matrix

Nanocomposite Fabrication Processes

Qiang Zhou, Assistant Professor, City University of Hong Kong,

Hong Kong, Hong Kong-PRC, [email protected],

Micheal De Cicco, Xiaochun Li, Shiyu Zhou, Li Zeng

Lightweight, high strength metal matrix nanocomposites (MMNCs) are promising materials. A uniform distribution of nanoparticles within the metal matrix is critical to its quality. However, a boundary effect often occurs where the nanoparticles tend to gather around the grain boundaries of the metal matrix. We propose a method for quantitatively assessing boundary effect in microstructure images of MMNC samples.

4 - Phase I Change Detection of Quality Profiles in Low-E Glass

Manufacturing Processes

Li Zeng, Assistant Professor, The University of Texas at Arlington,

500 West First Street, Arlington, TX, 76019,

United States of America, [email protected]

The concerns on energy and environment during the past decades have led to the manufacturing of low-emittance (low-E) glass. The current practice of quality monitoring in low-E glass manufacturing processes does not make a full use of the available quality profile measurements in these processes. In this study, we developed Bayesian approaches for the Phase I monitoring of quality profiles in these processes.

TC68

68- Suite 312- Hyatt

Financial Engineering & Regulations: From Credit

Risk to Energy

Sponsor: Financial Services Section

Sponsored Session

Chair: Tim Leung, Assistant Professor, Columbia University,

308A S. W. Mudd Building, 500 W. 120th Street, New York, NY, 10027,

United States of America, [email protected]

1 - A Comparison of the Original and Revised Basel Market Risk

Frameworks for Regulating Bank Capital

Alexandre Baptista, George Washington University, 2201 G Street

Northwest, Washington, DC, 20052, United States of America, [email protected], Gordon Alexander, Shu Yan

Risk management systems based on the original and revised Basel market risk frameworks allow the selection of trading portfolios with substantive tail risk.

However, a system based on the revised framework tends to be less effective in controlling tail risk. Importantly, the minimum capital requirements in the revised framework are much less likely to be wiped out by trading losses than those in the original framework. Thus, on balance, the revised framework improves upon the original framework.

2 - The Impact of Central Counterparty Design on Credit Default

Swap Market

Jinbeom Kim, Columbia University, 500 West 120th Street, New

York, NY, 10027, United States of America, [email protected]

Central clearing of credit default swaps (CDSs) through Central Counterparties

(CCPs) has been proposed as a tool for mitigating systemic risk and counterpart risk in CDS markets. We study the impact of the design of capital requirements of a CCP on the incentives for CDS clearing of the CDS inter-dealer market by deriving a unique Nash equilibrium of inter-dealer market CDS demands.

3 - Optimal Collateralization under Bilateral Default Risk

Daniel Bauer, Georgia State University, P.O. Box 4036, Atlanta,

GA, United States of America, [email protected], Luz R Sotomayor,

Enrico Biffis

We consider optimal collateralization in OTC transactions with bilateral default risk. In particular, we jointly solve for the transaction price and optimal collateral rules in a setting that takes into account the credit quality and funding needs of both parties. Our framework can be used for understanding Credit Support

Annex (CSA) pricing and Credit Value Adjustment (CVA) charges in structuring

OTC transactions. Our results provide insights on the key drivers of collateral rules.

4 - New Challenges in Electricity Price Modeling: Emissions,

Renewables and Market Coupling

Michael Coulon, Princeton University, ORFE Department,

Princeton, NJ, United States of America, [email protected]

Many electricity markets have recently undergone various fundamental changes linked to new regulations. These include the introduction of emissions markets, the growth of renewables and ongoing cross border integration (particularly in

Europe) via a mechanism called market coupling. Such key changes provide major obstacles for traditional reduced-form models of power price dynamics. We discuss these challenges and propose some alternative structural approaches to capturing these new features.

TC69

69- Suite 314- Hyatt

Advances in Risk Analysis

Contributed Session

Chair: Seong Dae KimAssistant Professor, University of Alaska

Anchorage, 3211 Providence Drive, University Center Room 155,

Anchorage, AK, 99508, United States of America, [email protected]

1 - Z-Theory and its Application to Risk Assessment and

Multi-Criteria Decision Making

Behnam Malakooti, Professor, Case Western Reserve University,

19000 Olin 611, Cleveland, OH, 44106, United States of America, [email protected]

This paper introduces Z-theory for risk assessment. The advantage of using Ztheory over classical expected utility theory is discussed and its robustness, simplicity, and applicability are explained. It is shown how Z-theory overcomes the disadvantages of Neumann and Morgenstern approach in modeling preferences. The theoretical foundation of Z-theory for decision problems under risk is also presented.

336

INFORMS Phoenix – 2012

2 - A Framework for Assessing Infrastructure Risk

Julia Phillips, Principal Infrastructure Analyst, Argonne National

Lab, 9700 S Cass Ave., Bldg 221, Argonne, IL, 60439, United States of America, [email protected], William Buehring, Jim Peerenboom,

Ron Whitfield, Lon Carlson, Gilbert Bassett

The functional form of Risk =T x V x C has encountered growing criticism, especially in the area of infrastructure risk analysis. We discuss the roots of such criticism and propose a framework in which the risk at a site for a given threattype is determined by consequences where consequences are represented by a multivariate random variable and the concepts of vulnerability and resilience in our approach become subsidiary/derived properties of the consequences distribution.

3 - Can Project Success be Accurately Predicted within Three

Months of Project Launch?

Rose Williams, Senior Technical Research Staff, IBM,

19 Skyline Drive, Hawthorne, 10532, United States of America, [email protected]

Statistical measures are used to initiate accelerated risk mitigation plans early in a software project’s lifecycle. Yet, how early in the lifecycle can an algorithm predict delivery results and expect reasonably accurate outcomes? Drawing upon a historical database of services projects spanning several years, we have developed an index that allows us to predict with reasonable certainty as early as 3 months into the development process — allowing us to plan for risk more effectively.

4 - Characterizing Unknown Unknowns when Facing

Natural Disasters

Seong Dae Kim, Assistant Professor, University of Alaska

Anchorage, 3211 Providence Drive, University Center Room 155,

Anchorage, AK, 99508, United States of America, [email protected]

When facing natural disasters, it’s hard to manage them because of their uncertain nature. Among different categories of uncertainty, unknown unknowns are known to be the vaguest in their characteristics and, often times, they become a surprising catastrophe due to the fact that they are almost impossible to manage. This paper reviews various ways to characterize uncertainties and seeks for characteristics of unknown unknowns in natural disasters to help make better decision in managing them.

5 - Risk Pooling in Performance-based Logistics

Guangyuan Yang, PhD Candidate, Erasmus Universiteit

Rotterdam, Burg. Oudlaan 50 H09-21, Rotterdam, 3062PA,

Netherlands, [email protected]

We assess the benefits of risk polling in PBL and its effects on aleatory uncertainties and epistemic uncertainties. Our results can enhance service performance and/or save costs in PBL by supporting logistics fulfillment of service contracts both at the initial phase (service contract sales) and during contract execution (with operational data).

TC70

70- Suite 316- Hyatt

Search and Social Media

Sponsor: Information Systems

Sponsored Session

Chair: Beibei Li, Carnegie Mellon University, 5000 Forbes Ave,

Pittsburgh, PA, 15213, United States of America, [email protected]

1 - Estimating Demand for Applications in the New Mobile

Economy

Sang-Pil Han, Assistant Professor, City University of Hong Kong,

Tat Chee Avenue, Kowloon, Hong Kong-ROC, [email protected], Anindya Ghose

We build a structural model of user demand for mobile applications. We use a panel dataset consisting of applications’ sales, prices, and characteristics from two leading app stores – Apple App Store and Google Android Market. Results show that demand increases with the filesize of apps, the age of apps, and the length of description. Counterfactual experiments show price discount strategies result in a greater increase of app demand in Apple App Store compared to Google Android

Market.

2 - Simultaneous or Sequential? Understanding the Drivers of

Search Strategies and Search in the U.S. Auto Insurance

Industry

Elisabeth Honka, Assistant Professor of Marketing, University of

Texas at Dallas, 800 W. Campbell Road, Richardson, TX, 75080,

United States of America, [email protected],

Pradeep Chintagunta

Whether consumers use simultaneous or sequential search strategies has long been a question of interest. The difficulty in settling this question stems from the fact that, in most data sets, the type of search is not identified so researchers have

TC71

to make an assumption regarding the specific search strategy being used. Using a unique data set on consumer shopping behavior in the U.S. auto insurance industry, we study the effects of making either assumption on the type of search.

3 - The Impact of Comments on Microblogging

Yingda Lu, Carnegie Mellon University,

5000 Forbes Avenue, Pittsburgh, PA, United States of America, [email protected], Yong Tan

In this paper, we investigate the impact of comments on user behavior on microblogging platform. We illustrate that comments of microbloggs could influence the way users decide whom to follow, whether to post microblogs. The comments can also influence the way new microblogs spread.

4 - Examining Limited Consumer Search with Social Media and

Product Search Engines

Beibei Li, Carnegie Mellon University,

5000 Forbes Ave., Pittsburgh, PA, 15213, United States of America, [email protected], Panagiotis Ipeirotis, Anindya Ghose

We propose a dynamic structural model of limited consumer search and validate it on a dataset of 1M online sessions for hotels. It allows us to monetize the search costs under social contexts. Our policy experiment shows that customized rankings can polarize the distribution of search intensity. Our model also indicates a lower price sensitivity than a static model would predict, implying that consumers pay a lot of attention to non-price factors during an online hotel search.

TC71

71- Suite 318- Hyatt

The Internet-Driven Economy

Sponsor: eBusiness

Sponsored Session

Chair: Raymond Sin, Assistant Professor, The University of Hong Kong,

Pokfulam, Hong Kong, Hong Kong - PRC, [email protected]

Co-Chair: Shun Ye, University of Maryland, RH Smith School of

Business, College Park, MD, United States of America, [email protected]

1 - Strategic Distribution and IT Investment Decisions in the Face of Digitization: Managerial Guideline

Nelson Granados, Pepperdine University, 15 Cedarlake, Irvine, CA,

92614, United States of America,

[email protected], John Mooney

The media, entertainment, and air travel industries are facing significant transformations due to the digitization of information and content. We use the empirical case study methodology to derive managerial guidelines for distribution strategy and related IT investments in the face of digitization. A tangible result of our work is a framework that enables firms to rationally decide whether to have a direct, joint-owned, or intermediated online distribution.

2 - Online Communities and Sustainability: An Empirical

Investigation

Dobin Yim, University of Maryland, 3300 Van Munching Hall,

College Park, MD, United States of America, [email protected]

Fostering consumer-driven energy conservation efforts has sprouted “green” online communities. We examine how online social networks facilitate environmental sustainability through the visibility of sustainable behaviors and group-induced norm. We show that slower growing groups perform better in energy reduction, and the strength of ties and group characteristics that individuals belong to are important predictors of prosocial behaviors.

3 - Overt Management Participation: Responding to Negative

Consumer Feedback in Online Review Forums

T. Ravichandran, Professor, Lally School of Management &

Technology, RPI, 110 Eighth Street, Troy, NY, 12180, United States of America, [email protected], Stacey Sharpe, Dongling Huang

In this paper we study the impact of managerial response strategies aimed at mitigating the potential implications of negative feedback generated by consumers in online review forumsófocusing specifically on factors related to the source and content of the managerial response. Using data collected from TripAdvisor.com on over 67000 consumer hotel reviews and corresponding responses from hotel administrators across two cities over ten years, we test our research hypotheses.

337

TC72

4 - Social Media, Conventional Media, and Stock Market

Performance: A Sentiment Analysis Approach

Wenjing Duan, Assistant Professor, The George Washington

University, 2201 G Street, NW, Washington, DC, 20052,

United States of America, [email protected], Qing Cao, Yang Yu

This study aims to investigate the effect of social media and conventional media, their relative importance, and their interrelatedness on firm stock market performances. We use a novel and large-scale dataset that features daily media content across various social media and conventional media outlets for 862 public traded firms. We apply the advanced sentiment analysis technique to analyze the overall sentiments of each media resource toward a specific company on the daily basis.

TC72

72- Suite 322- Hyatt

Exploration, Search, and Learning

Sponsor: Computational Stochastic Optimization

Sponsored Session

INFORMS Phoenix – 2012

Chair: Ilya Ryzhov, University of Maryland, 4322 Van Munching Hall,

College Park, MD, 21044, United States of America, [email protected]

1 - An Approximate Dynamic Programming Approach to

Sponsored Search

Andrew Mastin, PhD Student, Massachusetts Institute of

Technology, Room 32-D642, Cambridge, MA, 02139, United States of America, [email protected], Warren Powell, Patrick Jaillet

We present an approximate dynamic programming algorithm for online advertisement allocation in sponsored search (frequently referred to as the

Adwords problem). Using an approximation architecture based on piecewise linear separable value function approximations, we show empirical performance values greater than 99% of posterior solutions for problem instances where myopic polices are theoretically proven to perform poorly. We also show experimental results based on real search engine data.

2 - Controlling Greedy Sample Bias through Bias Corrrected

Q Learning

Donghun Lee, Princeton University, 35 Olden Street, Princeton,

NJ, 08540, United States of America, [email protected],

Warren Powell

We define non-negative bias caused by greedy sampling in exploitation steps of reinforcement learning using value estimation. We propose additive bias correction framework and apply it to Q learning to construct bias-corrected Q learning. Significant bias reduction and accelerated policy learning are demonstrated in Roulette simulation and intelligent battery control in power grid simulation.

3 - Learning to Learn

Benjamin Van Roy, Stanford University, 273 Packard,

Stanford, CA, United States of America, [email protected]

We will discuss the importance of learning to learn, and how this is a distinctive element of reinforcement learning relative to other areas of statistical learning.

We will then survey some relevant research and discuss recent work on an algorithm that efficiently learns to learn (and learns) in dynamic systems with arbitrarily large state spaces by combining optimistic exploration and value function generalization.

4 - The Robust Approach to Simulation Optimization

Ilya Ryzhov, University of Maryland, 4322 Van Munching Hall,

College Park, MD, 21044, United States of America, [email protected], Boris Defourny, Warren Powell

Simulation can be used to learn about a decision-maker’s environment and improve a future implementation decision (a system design, management policy, investment decision, etc.). The simulation optimization literature often assumes that decision-makers are risk-neutral. Robust optimization can be used to approach the problem from a risk-averse perspective. We formulate the riskaverse problem and propose a way to allocate the simulation budget based on an economic improvement criterion.

TC73

73- Suite 324- Hyatt

Production Planning

Contributed Session

Chair: Ali Kefeli, Operations Research Specialist, Kimberly-Clark Corp.,

1400 Holcomb Bridge Rd, Atlanta, GA, 30076, United States of

America, [email protected]

1 - Feedback Control of Machine Startup for Energy-Efficient

Manufacturing in Bernoulli Serial Lines

Guorong Chen, University of Wisconsin-Milwaukee, 3955 N

Murray Ave., apt 601, Milwaukee, WI, 53211, United States of

America, [email protected], Liang Zhang, Jorge Arinez,

Stephan Biller

Energy efficiency in manufacturing systems can be effectively improved by proper control of machine startup schedule. This paper derives closed-form formulas for performance measures in transient period, analyzes system theoretic properties, proposes effective controllers for machine startup schedule with examples, and thus illustrates the effectiveness of energy consumption reduction via proper machine startup control.

2 - Bi-Objective Optimization for Aggregate Planning with

Flexibility Requirements Profile

Edil Demirel, Teaching Assistant in Lee College of Engineering,

The Universirty of North Carolina at Charlotte, 9201 University

City Blvd., Cameron Building Room 291-A, Charlotte, NC, 28223,

United States of America, [email protected], Churlzu Lim,

Ertunga C. Ozelkan

Demand uncertainty causes changes in production plans, which create nervousness in organizations. Flexibility Requirements Profile (FRP) is an approach to reduce variability in production plans. This study proposes a biobjective optimization model for FRP-based aggregate planning, where the trade-off between plan stability and production cost is investigated through a design of experiments framework.

3 - A Genetic Algorithm for the Integrared Production, Inventory,

Distribution Routing Problem

Nihan Kabadayi, Research Asistant, Istanbul University, I.U.

Faculty of Business Administration, Avcilar, 34320, Turkey, [email protected], Timur Keskinturk

In this study, we aim to solve the integrated production, inventory,distribution routing (IPIDR) problem with using genetic algorithm (GA). The objective of the problem is to find optimal production quantity, customer delivery quantity and schedule in order to minimize the total system cost which is composed of production setup cost, inventory holding costs and distribution cost. We have developed a genetic algorithm to solve the IPIDR in an integrated manner.

4 - Testing the Quality of Personnel Rosters for Aircraft

Line Maintenance

Jorne Van den Bergh, PhD Student, HUBrussel, Warmoesberg 26,

Brussels, 1000, Belgium, [email protected],

Jeroen BeliÎn, Liesje De Boeck, Philippe De Bruecker,

Erik Demeulemeester

We present a simulation approach for the evaluation of personnel rosters of an aircraft maintenance company located at Brussels Airport. When constructing the rosters, the timing of the workloads was considered to be deterministic. The rosters are evaluated by simulating stochastic arrival times. We compare the realworld allocation method to other rules and analyze the performance of the different rosters by DEA.

5 - Duality of Capacity Constraints in Production Planning

Ali Kefeli, Operations Research Specialist, Kimberly-Clark Corp.,

1400 Holcomb Bridge Rd, Atlanta, GA, 30076,

United States of America, [email protected]

In this talk, we explore the dual behavior of the conventional method of capacity modeling. To gain insight we focus on the dual prices of capacity in a single stage single product production-inventory system. We derive closed form expressions of the dual variables and discuss the limitations of the modeling technique, in particular under which conditions they produce positive dual prices. We conclude with an exposition of an alternate capacity modeling technique with clearing functions.

338

Tuesday, 4:30pm - 6:00pm

TD01

01- West 101- CC

Integer Programming and Network Optimization

Sponsor: Optimization/Global Optimization

Sponsored Session

Chair: J. Cole Smith, Professor, University of Florida, P.O. Box 210020,

Gainesville, FL, 32611, United States of America, [email protected]

1 - Improve Power Grid with Controllable Reactance

Feng Pan, Los Alamos National Lab, Los Alamos National Lab,

Los Alamos, NM, 87545, United States of America, [email protected]

In a flexible AC transmission system, line reactance can be adjusted to improve a power grid’s performance. We use a linearized DC flow model to approximate power grid. With line reactance being decision variables, a bilevel program is developed to find an optimal setting of reactance on power lines. We discuss solution approaches and simulation results.

2 - Optimal Edge Search for an Immobile Hider with Imperfect

Detection Probability

Shantih Spanton, University of Florida, Gainesville, FL, 32611,

United States of America, [email protected], Joseph Geunes,

J. Cole Smith

We consider the problem of a single searcher traversing a network to find an immobile hider. The likelihood of finding the hider upon a particular edge decreases with each traversal the searcher makes upon the edge. Given a finite search time limit the problem maximizes the searcher’s probability of finding the hider on the network.

3 - An Algorithm for Bilevel Integer Interdiction

Yen Tang, University of Florida, 303 Weil Hall, Gainesville,

United States of America, [email protected], Jean-Philippe Richard,

J. Cole Smith

We propose a new algorithm for finding globally optimal solutions to a family of bilevel integer interdiction optimization in which the upper level minimizes the maximum of the objective of the lower level problem. We show that this algorithm, which proceeds by progressively building a suitable restriction of the lower level problem, is finitely convergent. We then present the results of a computational study aimed at evaluating the quality of our algorithm.

INFORMS Phoenix – 2012

TD03

2 - Modeling State-dependent Priorities of Malicious Agents

Sumitra Sri Bhashyam, PhD Candidate, London School of

Economics, Houghton Street, London, United Kingdom,

[email protected], Gilberto Montibeller

One way to anticipate terrorists’ malicious actions is to consider their values when modeling their decisions - such values will drive their choices. Using multiattribute utility theory we suggest modeling dynamic preferences of a malicious agent by incorporating state-dependent weights to account for preference change caused by exogenous triggers and representing the environment as a system dynamics model. We explore its use in the context of counter-terrorism analysis.

3 - Regulation Games Between Government and Competing

Companies: Oil Spills and Other Disasters

Jun Zhuang, Assistant Professor, University at Buffalo, SUNY, 403

Bell Hall, Buffalo, NY, 14260, United States of America, [email protected], May Cheung

Oil spills are a characteristic risk. There are safety regulations set to reduce the risk. Companies are balancing between safety efforts and production competition with other companies. We model and compare two games: a one-company game without competition and a two-company game with competition, both with the government as a regulator. Our results indicate that competition increases a company’s threshold for risk and therefore requires stricter government regulation.

TD02

02- West 102 A- CC

Joint Session DA/INFORMS Journals: Special Issue of Decision Analysis on Games and Decisions in

Reliability and Risk

Sponsor: Decision Analysis & INFORMS Journals Cluster

Sponsored Session

Chair: Jason Merrick, Professor, Virginia Commonwealth University,

Statistics & Operations Research, P.O. Box 843083, Richmond, VA,

23284, United States of America, [email protected]

Co-Chair: Fabrizio Ruggeri, Research Director, CNR-IMATI,

Via Bassini, 15, Milano, 20133, Italy, [email protected]

Co-Chair: Refik Soyer, Professor of Decision Sciences,

George Washington University, 2201 G Street, Washington, DC,

United States of America, [email protected]

1 - Adversarial Risk Analysis: The Somali Pirates Case

Jesus Rios, IBM, Yorktown Heights, NY, United States of America, [email protected], David Rios Insua, Juan Carlos Sevillano

We show how Adversarial Risk Analysis may cope with a current important security issue in relation with piracy in the Somali coasts. Specifically, we describe how to support the owner of a ship in managing risks from piracy in that area.

We illustrate how a sequential defend-attack-defend model can be used to formulate this decision problem and solve it for the ship owner. Our formulation models the pirates’ behavior through the analysis of how they could solve their decision problem.

TD03

03- West 102 B- CC

Resource Allocation and Investment Decisions

Sponsor: Decision Analysis

Sponsored Session

Chair: Juuso Liesio, Teaching Research Scientist, Aalto University,

Department of Mathematics and Systems Analysis, P.O. Box 11100,

Aalto, 00076, Finland, [email protected]

1 - Bayesian Evaluation and Selection Strategies in Portfolio

Decision Analysis

Eeva Vilkkumaa, Aalto University, P.O. Box 11100, Aalto, 00076,

Finland, [email protected], Juuso Liesio, Ahti Salo

In project portfolio selection, the projects’ future values are typically uncertain.

We show how the Bayesian modeling of these uncertainties increases the expected value of the portfolio, raises the expected number of optimal projects in the portfolio and decreases post-decision disappointment. We also show that it pays off to re-evaluate only those projects that can be re-evaluated relatively accurately and that have particularly uncertain initial value estimates near the selection threshold.

2 - Measurable Multiattribute Value Functions for Portfolio

Decision Analysis

Juuso Liesio, Teaching Research Scientist, Aalto University,

Department of Mathematics and Systems Analysis,

P.O. Box 11100, Aalto, 00076, Finland, [email protected]

The linear-additive value model (i.e. portfolio value is the sum of additive multiattribute project values) is widely employed to support multiobjective project portfolio selection even though the underlying preference assumptions do not often hold in practice. We relax these assumptions and derive a more general class of multilinear portfolio value functions. Furthermore, we develop techniques to elicit these value functions and algorithms to identify the optimal project portfolio.

3 - Resource Based Path Dependency and Risk Aversion

Janne Kettunen, Assistant Professor, Haskayne School of Business,

2500 University Drive N.W., Calgary, AB, Canada, [email protected], Derek Bunn

We provide reasons for the origin of firm’s resource based heterogeneity. Given heterogeneous existing resources, we show that it is optimal for a firm with hedgeable existing assets to signal that it is more risk averse. Also, we show that steps toward asset diversification are path dependent as a function of time. For policy makers, our results suggest that if resource based path dependency is not considered policy impacts can be significantly different than what expected.

4 - Communicating Decision Analytic Results to Decision Makers

Patrick Noonan, Professor in the Practice of Decision & Info.

Analysis, Emory University, 1300 Clifton Road NE, Goizueta

Business School, Atlanta, GA, 30322, United States of America, [email protected], Jeff Keisler

Good analysis is not enough to define success in a decision analysis (DA) engagement. Good communication of analytical results helps stakeholders to understand, accept and implement recommendations. DA itself provides several useful frameworks for approaching the decision of how to communicate.

Standard communication techniques in DA engagements can be improved further by drawing on insights from other areas of management practice.

339

TD04

INFORMS Phoenix – 2012

TD04

04- West 102 C- CC

Data Envelopment Analysis IV

Cluster: Data Envelopment Analysis

Invited Session

Chair: Andrew Johnson, Texas A & M University, 4033 ETB, College

Station, TX, United States of America, [email protected]

1 - A DEA Study of Sustainable Energy Efficiency in New Zealand

Julie Harrison, Senior Lecturer, University of Auckland,

Department of Accounting & Finance, 12 Grafton Road, Auckland,

New Zealand, [email protected], Paul Rouse,

Jennifer Chen

Our study examines macro-level energy efficiency trends in New Zealand from

1990 to 2009. During this period the NZ government implemented a number of sustainability initiatives, including the adoption of the Kyoto Protocol. We use a

DEA model that combines greenhouse gas emissions data and total factor energy efficiency measures. The results show improvements in energy efficiency, with the largest improvements concurrent with the introduction of major government initiatives on sustainability.

2 - Assessment and Identification of Efficient Strategies for

Prevention of Obesity

Nasim Sabounchi, Virginia Technical University,

7054 Haycock Road, Room 433, Falls Church, VA, 22044,

United States of America, [email protected], Kostas Triantis

Due to the increasing obesity trend, we need to identify the most efficient and sustainable intervention strategies in turning the obesity trend. Our objective is to study how efficiency in implementing the obesity interventions, affects resources required, intervention capabilities, and obesity mitigation. We adopt a DEA approach to find appropriate model specifications, define the inputs, outputs and outcomes for the transformation process and present insights in evaluating obesity policies.

3 - Supplier Evaluation and Resilient Base Identification in

Presence of Supply Disruptions

Anthony Ross, Professor, University of Wisconsin Milwaukee,

P.O. Box 742, Milwaukee, WI, 53201, United States of America, [email protected], Wanxi Li, Kaan Kuzu

This study investigates suppliers’ performance under performance outcome uncertainty, and further explores the resilience of supplier’s performance in disrupted environment. We propose a framework to evaluate suppliers’ performance capabilities in the context disruption occurrences. Such a framework may assist supply managers in proactively mitigating the potential impact of the disruption by understanding capabilities imbedded within the supply base.

Several managerial insights are discussed.

4 - Allocative Efficiency in an Oligopolistic Market: An Example of

Rational Inefficiency

Andrew Johnson, Texas A & M University, 4033 ETB, College

Station, TX, United States of America, [email protected],

Chia-Yen Lee

This paper models the production possibility set and the inverse demand function to identify a Nash equilibrium and improvement targets which may not be on the production frontier. This behavior is referred to as rational inefficiency because the firm reduces its productivity levels in order to increase profits. For a general multiple input/output production process which allows a firm to adjust variable input and output levels, the existence and the uniqueness of the Nash equilibrium is proven.

TD05

05- West 103 A- CC

Technometrics Invited Session New Directions in

Statistical Process Control

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Roshan Vengazhiyil, Georgia Institute of Technology,

765 Ferst Drive, NW, Atlanta, GA, 30332, United States of America, [email protected]

1 - On Nonparametric Statistical Process Control of

Univariate Processes

Peihua Qiu, Professor, University of Minnesota,

224 Church Street SE, 313 Ford Hall, Minneapolis, MN, 55455,

United States of America, [email protected], Zhonghua Li

This paper considers statistical process control (SPC) of univariate processes when the parametric form of the process distribution is unavailable. We propose a new framework for constructing nonparametric control charts, by first categorizing observed data and then applying categorical data analysis methods to SPC. Some new nonparametric control charts are proposed and they are compared with several representative existing control charts in various cases.

2 - A LASSO-Based Diagnostic Framework for Multivariate

Statistical Process Control

Wei Jiang, Professor, Shanghai Jiaotong University,

Antai College of Economics & Management, Shanghai, China, [email protected], Fugee Tsung, Changliang Zou

Accurate fault diagnosis of responsible factors has become increasingly critical in a variety of applications. Conventional statistical process control (SPC) methods are often computationally expensive. We frame fault isolation as a two-sample variable selection problem to provide a unified diagnosis framework based on

Bayesian information criterion (BIC). A practical LASSO-based diagnostic procedure is proposed which combines BIC with the popular adaptive LASSO variable selection method.

3 - Discussion

Arthur Yeh, Professor, Bowling Green State University,

Wooster Street, Ann Arbor, OH, 43403, United States of America, [email protected]

I will discuss the presented papers and give general thoughts about the future of statistical process control.

4 - Discussion

Daniel Apley, NorthWestern University, 2145 Sheridan Road,

Evanston, IL, 60208, United States of America, [email protected]

I will discuss the presented papers and give general thoughts about the future of

SPC.

TD06

06- West 103 B- CC

Agent-based Simulation of Complex Adaptive

Systems

Sponsor: Simulation

Sponsored Session

Chair: Moeed Haghnevis, Graduate Research and Teaching Associate,

Arizona State University, Tempe, AZ, 85287-8809,

United States of America, [email protected]

1 - Scalability of Modeling Driver’s Behavior under the Extended

Belief-desire-intention Framework

Sojung Kim, The University of Arizona, 1127 E. James E, Rogers

Way, Tucson, AZ, 85721, United States of America, [email protected], Hui Xi, Young-Jun Son

While the extended Belief-Desire-Intention (BDI) framework allows us to mimic realistic driver behaviors, it is computational demanding when applied to a large traffic network. This paper aims at analyzing the relationship between model scalability and network complexity. By comparing the return of investment of employing different modeling methodologies, we intend to find the scenarios in which the extended BDI framework can achieve high performance in terms of both accuracy and efficiency.

2 - Agent-based Simulation of Healthcare Systems with Real-time

Data Feeding

Xueping Li, University of Tennessee, 408 East Stadium Hall,

Knoxville, TN, 37996, United States of America,

[email protected], Tom Berg, Joe Wilck

Real-time location services (RTLS), such as Wifi-based RFIDs, have attracted increasing interest in healthcare systems. We present the simulation models that we developed for the Y-12 Occupational Health Services (OHS) clinic in which we use agents to fetch sensor data and feed the models in real time. A management system was developed to assist the end-users with a dashboard system to provide real-time system status. Optimization and predictive modeling functions were integrated in the system.

3 - An Agent-based Dynamic Laboratory for Electricity

Consumption Complex System

Moeed Haghnevis, Graduate Research and Teaching Associate,

Arizona State University, Tempe, AZ, 85287-8809, United States of

America, [email protected], Dieter Armbruster, Ronald Askin

The objective of this research is to construct an agent-based platform and toolkit that creates a dynamic laboratory to study properties and behavior of electric power grid supply and consumption networks and related engineered complex adaptive systems. In addition to developing an understanding of structural behavior, social networking interactions and consumer behaviors are integrated to manage demand in an environment with bi-directional information flow and variable supply.

340

INFORMS Phoenix – 2012

TD07

07- West 104 A- CC

Forecasting

Contributed Session

Chair: Martin Braun, Intel Corp, 5000 W. Chandler Blvd, Chandler, AZ,

85295, United States of America, [email protected]

1 - Determining the Right Buffer Strategy during Red River Floods,

ND, MN

Luke Holt, Graduate Assistant, North Dakota State University:

Transportation and Logistics Program, 7080 148th Ave., NE,

Grafton, ND, 58237, United States of America, [email protected], Joseph Szmerekovsky, Jiang Zhang, Lihui Bai

The Red River of the North has created flooding problems to North Dakota and

Minnesota for many years. water levels are highly unpredictable from year to year. Planners face the challenge to determine what buffer strategy provides appropriate protection. Dynamic programming is used to determine the best buffer strategy.

2 - Scenarios as Support Tools for Judgmental Forecasts

M.Sinan Gonul, Assistant Professor, METU, Department of

Business Administration, Middle East Technical University,

Ankara, 06800, Turkey, [email protected], Dilek Onkal,

K.Zeynep Sayim

This study examines the effects of providing scenarios on individual and groupbased judgmental predictions. In the individual forecasting stage, participants are presented with: i) no scenarios, ii) best-case scenarios, iii) worst-case scenarios, iv) both best-case and worst-case scenarios. For the group forecasting stage, members of dyads may be assigned: a) no scenarios, b) different scenarios, c) both scenarios. Findings are discussed and implications for future research are given.

3 - Can Oil Prices be a Proxy for Consumer Sentiment?

Michelle Tuveson, University of Cambridge Judge Business School,

Trumpington Street, Cambridge, CB2 1AG, United Kingdom, [email protected], Daniel Ralph

Efficient Market Hypothesis states that income and prices are related to spending irrespective of the behavioral dynamics contributing to economic activity.

Research around consumer sentiment has largely focused on its use as a predictive measure on economic forecasts. We pose that consumer sentiment measures are informative metrics for conveying general mood within the macroeconomy. Oil price data is shown to be a significant comparative proxy using machine learning applications.

4 - Applying Adaptive EWMA Tuning Methods to the Demand

Forecast Problem

Martin Braun, Intel Corp, 5000 W. Chandler Blvd, Chandler, AZ,

85295, United States of America, [email protected]

Applying EWMA and other filtering methods to demand forecast error for bias and noise estimation can be difficult due to noise characteristics varying with time. A few adaptive methods of tuning EWMA filters are applied successfully in the case of continuous and intermittent demand forecasting problems.

TD08

08- West 104 B- CC

Joint Session PPSN/HAS: Health Care Policy Analysis

Sponsor: Public Programs, Service and Needs & Health

Applications Society

Sponsored Session

Chair: David Hutton, University of Michigan, 1415 Washington

Heights, SPH II: M3525, Ann Arbor, MI, 48109, United States of

America, [email protected]

1 - Cost-effectiveness of Risk-factor Guided and Universal

Screening for Chronic Hepatitis C

Shan Liu, Stanford University, Stanford, CA, 94305,

United States of America, [email protected], Lauren Cipriano,

Mark Holodniy, Jeremy Goldhaber-Fiebert

U.S. guidelines conflict over population screening for chronic hepatitis C. We assessed the cost-effectiveness of universal and risk-factor guided screening for asymptomatic adults followed by one of three treatment strategies. Statistical analyses of the National Health and Nutrition Examination Survey estimated sex-, race-, age-specific risk factor prevalence and mortality risks. Universal screening is likely cost-effective.

TD09

2 - Optimal Health Program Intervention with Information

Acquisition

Lauren Cipriano, Stanford University, Huang Engineering Center,

475 Via Ortega, Stanford, CA, 94305, United States of America, [email protected], Thomas Weber

We consider a cohort-specific health-intervention decision as an optimal stopping problem. Across successive periods a given age cohort’s imperfectly observable disease prevalence evolves randomly. To inform the stopping decision it is possible to collect a costly sample, the size of which is subject to optimization. We determine an optimal program-continuation and information-collection policy and, as application, propose an optimal testing program for hepatitis C at routine medical visits.

3 - Optimizing Age-specific Screening Strategy for Genital

Chlamydia Infections in Women

Nan Kong, Assistant Professor, Purdue University, 206 S. Martin

Jischke Drive, West Lafayette, IN, United States of America, [email protected], Yu Teng, Wanzhu Tu

Genital chlamydia infection is a common sexually transmitted disease. Since most infected individuals are asymptomatic and the incidence rate differs by age, we conduct a cost-effectiveness analysis on age-specific screening strategies. We present a discrete-time optimal control problem with cumulative care cost and terminal prevalence as the objectives. We solve the problem with respect to several age-specific screening strategies and address the computational efficiency issue.

4 - Analyzing the Impact of Government Participation on the U.S.

Healthcare Financing System

Zhen Zhu, Purdue University, 315 N. Grant Street, West Lafayette,

IN, United States of America, [email protected], Andrew Liu,

Nan Kong

In healthcare financing, over-utilization and insurance overhead are two major structural costs that do not contribute to well-being. We consider a mechanism design problem, for which we formulate as a leader-follower game. The government is the leader who intends to minimize the total structure cost, and the representative private insurer is the follower who intends to maximize its profit. Three mechanisms are studied and numerical analysis is conducted based on the RAND HIE data.

TD09

09- West 105 A- CC

Preference Learning II

Sponsor: Multiple Criteria Decision Making

Sponsored Session

Chair: Salvatore Greco, Professor, University of Catania, Corso Italia 55,

Catania, 95129, Italy, [email protected]

1 - Predicting Manhole Events in New York City

Cynthia Rudin, Assistant Professor, Massachusetts Institute of

Technology, 77 Massachusetts Avenue, Cambridge, MA,

United States of America, [email protected]

I will describe the Columbia/Con Edison Manhole Events Project, the goal of which is to predict electrical failures, namely manhole fires and explosions in New

York City. I will discuss the data mining process by which we transformed extremely raw historical Con Edison data into a ranking model that predicts manhole vulnerability. Our ranked lists are currently assisting with the prioritization of future inspections and repairs in Manhattan, Brooklyn, and the

Bronx.

2 - Interactive Evolutionary Multicriteria Scheduling

Jon Marquis, [email protected], John Fowler, Esma Gel,

Murat Koksalan, Jyrki Wallenius, Pekka Korhonen

We show an approach to multicriteria parallel machine job scheduling that uses an interactive genetic algorithm to assign jobs to machines and a scheduling heuristic to order those jobs. The algorithm interacts with the decision maker to elicit preference information and guide the algorithm’s search. We demonstrate the algorithm’s effectiveness for two and three criteria problems.

341

TD10

INFORMS Phoenix – 2012

TD10

10- West 105 B- CC

New Models and Algorithms for Discrete

Optimization with Uncertainty

Sponsor: Optimization/Stochastic Programming

Sponsored Session

Chair: Bo Zeng, Assistant Professor, University of South Florida, 4202 E

Fowler Ave ENB118, Tampa, FL, 33620, United States of America, [email protected]

1 - An Approximation Method to Solve Two-stage

Stochastic Program

Yu An, University of South Florida, 3602 Jefferson Commons

Drive Apt302, Tampa Fl 33613, tampa, Fl, 33613,

United States of America, [email protected], Bo Zeng, Long Zhao

Two-stage stochastic program is known computational challenging as a large number of scenarios need to consider. In this research, we consider to approximate stochastic programs, especially stochastic MIP problems, by solving a series of two-stage robust optimization problems with disjoint uncertainty sets.

Preliminary computational results will be presented to evaluate the performance of this approximation strategy.

2 - A Benders’ Decomposition Algorithm for

Chance-Constrained MIPs

Ludwig Kuznia, Walt Disney World, 1375 E. Buena Vista, Lake

Buena Vista, United States of America, [email protected],

Bo Zeng

In this presentation, an innovative application of Benders’ decomposition is presented as a solution method for chance-constrained MIPs. Roughly speaking, this can be viewed as a two-phase approach. In the first phase, valid cutting planes are found by solving a relaxed problem. This information is used to seed the Benders’ algorithm in the second phase.

3 - Branch-and-Cut Algorithms for Chance-constrained

Packing Problem

Yongjia Song, University of Wisconsin-Madison, 1513 University

Avenue, 3241 Mechanical Engineering Building, Madison,

United States of America, [email protected], James Luedtke,

Simge Kucukyavuz

We study the chance-constrained packing problem with random item weights.

We propose new inequalities based on probabilistic covers and strategies for lifting them. We also present a decomposed projection from the extended MIP formulation.

4 - A Two-stage Stochastic Programming Model for Phlebotomist

Scheduling in Hospital Laboratories

Laquanda Leaven, PhD Candidate, North Carolina A&T State

University, 1601 E. Market Street, Greensboro, NC, 27401,

United States of America, [email protected], Xiuli Qu

Laboratory services in healthcare systems play a vital role in inpatient care. Most hospital laboratories face the challenge to reduce cost and improve service quality.

In our study, a Two-Stage Stochastic Programming Model was formulated to determine better phlebotomist schedules for the preanalytical stage of the testing process in hospital laboratories. The objective is to balance the workload among shifts and the workload among phlebotomists in each shift, which result in cost reductions.

2 - Hot Started NLP Solvers: Applications and

Numerical Experiments

Travis Johnson, Northwestern University, 2145 Sheridan Road,

Engineering Science/Applied Math, Evanston, IL, 60208, United

States of America, [email protected], Andreas Waechter

We discuss an algorithm for quickly solving a sequence of closely related nonlinear continuous optimization problems, e.g., during the solution of MINLPs or optimal control problems. The algorithm exploits hot-starts of active-set QP solvers, avoiding the refactorization of KKT matrices. This presentation examines the numerical performance of the method.

3 - Scalable Dynamic Optimization

Victor Zavala, Argonne National Laboratory, Mathematics &

Computer Science Division, Chicago, IL, 60439,

United States of America, [email protected]

We discuss scalability issues arising in dynamic optimization and present potential strategies to avoid them. In particular, we present alternatives to enable fast active-set detection and matrix-free implementations.

TD12

12- West 106 A- CC

Algorithmic Learning Theory

Sponsor: Optimization/Integer Programming

Sponsored Session

Chair: Karthekeyan Chandrasekaran, Georgia Institute of Technology,

266 Ferst Drive, KACB 2116, Atlanta, GA, 30332,

United States of America, [email protected]

1 - Structure from Local Optima: Learning Subspace Juntas via

Higher Order PCA

Ying Xiao, Georgia Institute of Technology, 266 Ferst Drive,

Atlanta, GA, 30332, United States of America, [email protected], Santosh Vempala

Our result is an algorithm for generalized Independent Component Analysis that uses local optima of high moments to recover the component subspaces of a product distribution. If one component is k-dimensional while the other is “noise” modeled as an (n-k)-dimensional Gaussian, the algorithm is polynomial in n. We use this to learn a k-subspace junta, a labeling function determined by an unknown k-dimensional subspace, which is a generalization of learning a junta.

2 - Distributed Non-stochastic Experts

Varun Kanade, Harvard University, 33 Oxford Street #138,

Cambridge, MA, 02138, United States of America, [email protected], Zhenming Liu, Bozidar Radunovic

We consider the non-stochastic experts problem in a distributed system with k sites and one co-ordinator node that communicates with all sites. On any round a site has to choose an action: the tradeoff being between using stale information or using communication to get latest information. Recent techniques in streaming and distributed optimization literature do not yield non-trivial algorithms. We give a novel algorithm that simultaneously achieves good regret bounds and sublinear communication.

3 - Stability Yields a PTAS for k-Median and k-Means Clustering

Pranjal Awasthi, CMU, 5000 Forbes Avenue, Pittsburgh, PA,

United States of America, [email protected], Avrim Blum,

Or Sheffet

We present a new notion of clustering stability under which one can design polynomial time approximation schemes for k-means and k-median clustering objectives. Our result improves upon previous work of [Ostrovsky’06] and

[Balcan, Blum, Gupta’09].

TD11

11- West 105 C- CC

Algorithms for Nonlinear Optimization and Optimal

Control

Sponsor: Optimization/Nonlinear Programming

Sponsored Session

Chair: Andreas Waechter, Northwestern University, 2145 Sheridan

Road, Evanston, IL, 60208, United States of America,

[email protected]

1 - Hot Started NLP Solvers: Algorithms

Andreas Waechter, Northwestern University, 2145 Sheridan Road,

Evanston, IL, 60208, United States of America,

[email protected], Travis Johnson

We discuss an algorithm for quickly solving a sequence of closely related nonlinear continuous optimization problems, e.g., during the solution of MINLPs or optimal control problems. The algorithm exploits hot-starts of active-set QP solvers, avoiding the refactorization of KKT matrices. This presentation describes the details of the algorithm.

342

INFORMS Phoenix – 2012

TD13

13- West 106 B- CC

Strong Formulations and Cutting Planes for MIP

Sponsor: Optimization/Integer Programming

Sponsored Session

Chair: Ismael de Farias, Texas Tech University, Department of Ind. Eng,

Lubbock, TX, 79409-3061, United States of America, [email protected]

1 - Computations with Decomposition Algorithms using Gomory

Cuts for 2-Stage Stochastic IPs

Dinakar Gade, Ohio State University, 1971 Neil Ave., Columbus,

United States of America, [email protected], Simge Kucukyavuz,

Suvrajeet Sen

We consider a class of two-stage stochastic integer programs with binary variables in the first stage and general integer variables in the second stage. We develop a finitely convergent decomposition scheme utilizing Gomory cuts to obtain iteratively tighter approximations of the second stage integer programs. We present computational experiments that illustrate the use of these cuts both within a pure cutting plane, as well as a branch-and-cut based decomposition algorithm.

2 - A Computational Study of Linear and Integer Optimization with

Multiple-Choice Restrictions

Ismael de Farias, Texas Tech University, Department of Ind. Eng,

Lubbock, TX, 79409-3061, United States of America, [email protected], Ernee Kozyreff, Ming Zhao

A wide variety of applications require that certain variables have value zero in a feasible solution when other specific variables are nonzero. Such constraints are called multiple-choice. We present new inequalities for solving multiple-choice optimization through branch-and-cut, and an extensive computational experience that reveals new and surprising facts about these problems, their formulations, and the cutting planes.

3 - Strong Formulations for Convex Functions Over

Nonconvex Sets

Daniel Bienstock, Columbia University, Department of IEOR,

New York, NY, United States of America, [email protected],

Alex Michalka

We describe continuing work for obtaining the convex hull of sets representing quadratic polynomials over nonconvex sets, such as the complement of a polyhedron, the complement of a paraboloid, or the complement of a union of polyhedra and paraboloids.

4 - Column Generation Stabilization using Dual Smoothing:

Theory and Practice

Francois Vanderbeck, Université de Bordeaux, Institut de

Mathématiques, Bordeaux, France, [email protected],

R. Sadykov, E. Uchoa, J. Han, P. Pesneau, A. Pessoa

Cutting plane algorithmic strategies translate into stabilization procedures for column generation. Some techniques independently developed in the literature are in fact dual equivalent of one another. The link between in-out separation and dual price smoothing helps to establish convergence proves and effective smoothing auto-regulating strategies that avoids the need for parameter tuning.

We develop dual smoothing and provide bench- mark against existing stabilization techniques.

TD16

2 - Vehicle Dispatching for New Automated Container Terminal

Ek Peng Chew, Associate Professor, National University of

Singapore, 10 Kent Ridge Crescent, Singapore, Singapore, [email protected], Yanhua Xu, Loo Hay Lee

In this talk, we will present a new automated container terminal (ACT) concept.

The disptaching of transport vehicles in this ACT is challenging and we porpose some algorithms to provide efficient solutions for this problem.

3 - Bike-Sharing System Design and Scheduling

Mabel Chou, NUS Business School, Level 8, Biz1, Singapore,

Singapore, [email protected], Meilin Zhang

We develop practical OR models to support decision making in the design and scheduling of public car-sharing or bicycle-sharing systems. We develop a network flow model and discuss how to design the system so that the scheduling of the re-distribution can be more cost effective.

4 - Scheduling of Non-resumable Deteriorating Jobs on a Single

Machine with Multi-period Maintenances

Hui-Chih Hung, Assistant Professor, National Chiao Tung

University, MB103, 1001, University Rd., Hsin-Chu, 300, Taiwan-

ROC, [email protected], Hsi-Mei Hsu, Hsun-Wen Hsu

We consider a scheduling problem of non-resumable deteriorating jobs on a single machine with multi-period maintenances. The objective is to minimize the makespan. We show that the problem with 2-period maintenances can be reformulated as a double knapsack problem and is NP-hard. A fast heuristic is proposed to find a near optimal solution subject to arbitrary relative errors.

Numerical studies are implemented to verify the corresponding computational time subject to the given relative errors.

TD15

15- West 202- CC

Software Demonstration

Invited Session

1 - IBM ILOG CPLEX Optimization Studio

Ferenc Katai, IBM, 1681 Rue des Dolines, Les Taussiniers HB2,

Valbonne 06560, France, [email protected], Ed Klotz

Come learn how the latest integrated development environment enhancements combined with engine performance improvements and new features help mathematicians, OR experts, academicians and business analysts leverage advanced analytics based on CPLEX optimization to make fact-based decisions in rapidly changing markets. A live demonstration of CPLEX Optimization Studio will be presented by experts from the largest staff of advanced analytics and optimization professionals of any company in the world.

2 - Tableau Software - Stop Wrestling with Your Data-Start

Exploring it with Tableau

Jonah Kim, Tableau Software, 837 N. 34th St., #400, Seattle WA,

United States of America, [email protected]

Tableau Desktop is based on breakthrough technology from Stanford University that lets you drag and drop to analyze data. You can connect to data in a few clicks, then visualize and create interactive dashboards with a few more. Tableau

Software transforms stubborn databases and spreadsheets into sources for easy investigations. It’s so easy to use that any Excel user can learn it. Answer questions as fast as you can think of them.

TD14

14- West 106 C- CC

Scheduling Models and Applications

Cluster: Scheduling and Project Management

Invited Session

Chair: Hui-Chih Hung, Assistant Professor, National Chiao Tung

University, MB103, 1001, University Rd., Hsin-Chu, 300, Taiwan -

ROC, [email protected]

1 - Scheduling Subject to Fixed Job Sequences

Bertrand Lin, Professor, National Chiao Tung University,

1001 Tahsueh Road, Hsinchu, 300, Taiwan-ROC, [email protected], Feng-Jang Hwang

In most scheduling problems, if job/operation sequences on the machines are known, then the best schedules corresponding to the given sequences can be easily derived. For some problems, determining such schedules is however nontrivial. This talk discusses several scheduling problems with this aspect by presenting dynamic programming algorithms and NP-hardness proofs. Some open questions concerning complexity status will be addressed, too.

TD16

16- West 207- CC

Particle Filtering

Cluster: Tutorials

Invited Session

Chair: John Birge, Professor, University of Chicago, Chicago, IL,

United States of America, [email protected]

1 - Particle Filtering

John Birge, Professor, University of Chicago, Chicago, IL,

United States of America, [email protected]

Particle methods provide a robust methodology for estimation and prediction.

They can also apply to traditional operations research areas by, for example, providing an effective alternative for constructing Monte Carlo simulations of various systems and for dynamic optimization. This tutorial explains the fundamental ideas behind particle methods and how to use them in typical operations research contexts.

343

TD17

INFORMS Phoenix – 2012

TD17

17- West 208 B- CC

Operations Management, Finance Interface

Contributed Session

Chair: Thomas Yeung, Associate Professor, Ecole des Mines, 4 rue

Alfred Kastler BP 20722, Nantes, 44307, France, [email protected]

1 - Inventory and Working Capital Management

Kasper van der Vliet, PhD candiate, Eindhoven University of

Technology, Den Dolech 2, Eindhoven, 5612 AZ, Netherlands, [email protected], Matthew Reindorp, Jan Fransoo

Inventory management and working capital management are traditionally separate domains, and thus theory often neglects their interaction. We explicitly study a system where inventory can be financed by cash, reverse factoring, or borrowing. This allows us to explore the interaction between inventory policy, trade credit terms, and cash management thresholds.

2 - A Robust Portfolio Approach to System-of-Systems

Network Architectures

Navindran Davendralingam, Post Doctoral Researcher, Purdue

University, 701 W. Stadium Avenue, West Lafayette, United States of America, [email protected], Daniel DeLaurentis

This research presents a novel robust network framework to architecting a

System-of-Systems (SoS). Hierarchies of interdependent systems are modeled as nodes on a network that work cohesively to fulfill overarching capability objectives. Inter-nodal performance and modes of risks associated with connectivity under varied conditions of operation are addressed within the context of a robust mean-variance investment problem. A Littoral Combat Ship

(LCS) sample case demonstrates the method.

3 - Association between Automotive Recalls and Performance

Manpreet Hora, Assistant Professor, Georgia Institute of

Technology, College of Management, 800 West Peachtree St. NW,

Atlanta, GA, 30332, United States of America,

[email protected], Nah Lee, Vinod Singhal

Studies have shown mixed results (negative or null) in investigating the impact of product recalls on the stock value of recalling firms. We collect announcements related to automotive recalls to empirically examine their association with shareholder value. We also investigate the association between different product characteristics of the vehicles recalled and shareholder value.

4 - Information Acquisition of New Technology for Maintenance and Replacement Decisions

Thomas Yeung, Associate Professor, Ecole des Mines, 4 Rue Alfred

Kastler BP 20722, Nantes, 44307, France, [email protected],

Phuong Khan Nguyen Thi, Bruno Castanier

An important mission of managers is to determine the maintenance and replacement investment plan for equipment under technological evolution. The arrival time and profitability of new technology are uncertain; however information on this process may be acquired. We formulate a partially observable

Markov decisions process to decide whether to obtain additional information as well as the action for the asset (wait, maintain, replace). We provide structural properties and numerical examples.

TD18

18- West 208 A- CC

New Developments in the PICO/PEBBL Parallel

Branch and Bound Solver

Sponsor: Optimization/Computational Optimization and Software

Sponsored Session

Chair: Cynthia Phillips, Senior Scientist, Sandia National Laboratories,

Mail Stop 1326, P.O. Box 5800, Albuquerque, NM, 87185-1326,

United States of America, [email protected]

1 - Applying PEBBL Parallel Branch and Bound to Protein/Peptide

Docking and Maximum Monomial Agreement

Jonathan Eckstein, Rutgers University, 640 Bartholomew Road,

Piscataway, NJ, 08854, United States of America, [email protected], Cynthia Phillips, William E. Hart

We discuss applying the PEBBL parallel branch-and-bound library to a quadratic semi-assignment approximation of the protein-peptide docking problem, with a strengthened bounding procedure having the same complexity as prior bounds.

This application exhibits strongly and reproducibly superlinear speedups definitively attributed to processor cache behavior. We also describe applying

PEBBL to maximum monomial agreement, an problem arising from machine learning.

2 - Computing Certificates for Integer Programs

Ojas Parekh, Senior Member Technical Staff, Sandia National

Laboratories, MS 1326, Albuquerque, NM, 87185,

United States of America, [email protected], Cynthia Phillips,

Robert Carr, Harvey Greenberg

A certificate for a computation is data that allows an independent program to verify that its output is correct. A canonical example is the primal/dual certificate for a linear program, requiring only two matrix evaluations to prove the optimality of any solution. A certificate for an IP branch-and-bound computation must prove that each branching operation, added cut, and fathoming operation is correct. We will discuss what this entails in the context of PICO, our massively parallel IP solver.

3 - PICO’s New Hierarchical Branch-and-bound System for

Massively Parallel IP

Cynthia Phillips, Senior Scientist, Sandia National Laboratories,

Mail Stop 1326, P.O. Box 5800, Albuquerque, NM, 87185-1326,

United States of America, [email protected], Jonathan Eckstein,

Ojas Parekh, John D. Siirola, Jean-Paul Watson

We present the design, implementation, and large-scale parallel computational results for a new capability in the PICO (Parallel Integer and Combinatorial

Optimizer) massively-parallel mixed-integer programming (IP) solver. We leverage the PICO ramp up system for automatic IP decomposition and carefully manage runtime conditions to effectively run arbitrary black-box IP solvers on massively parallel systems. Our computational results use a Sandia National

Laboratories’ 20,000-core system.

TD19

19- West 211 A- CC

Resource Management in Health Care

Contributed Session

Chair: Mihiro Sasaki, Prof., Nanzan University, 27 Seirei, Seto,

489-0863, Japan, [email protected]

1 - Integral Resource Capacity Planning for Inpatient Care Services

Aleida Braaksma, Academic Medical Center Amsterdam &

University of Twente, Drienerlolaan 5, Enschede, 7500 AE,

Netherlands, [email protected], Nikky Kortbeek,

Ferry Smeenk, Richard J. Boucherie, Piet J.M. Bakker

The challenge in decision making for inpatient care delivery is to guarantee care from appropriately skilled nurses and required equipment for patients, while making efficient use of resources. We present a stochastic analytical approach to predict bed census on nursing wards by hour, as a function of the Master Surgical

Schedule and arrival patterns of emergency patients. Based on the hourly census predictions, flexible nurse staffing methods are proposed.

2 - Using On-call Providers for Emergency Departments

Subhamoy Ganguly, Leeds School of Business, University of

Colorado at Boulder, Boulder, United States of America,

[email protected], Stephen Lawrence,

Mark Prather

Medical emergency departments (EDs) face stochastic demand in terms of patient arrivals and their acuity levels. To meet high service-levels while keeping staffing costs under control, EDs often have providers on call, who are summoned to handle any unforeseen surge in demand. Using empirical data collected from three active EDs, we analyze how on-call providers can be effectively used in conjunction with a base staffing schedule to deliver high service-level with minimal staffing costs.

3 - A Decision Support System for Case Mix and Capacity Planning at Jessa Hospital

Guoxuan Ma, Faculty of Business and Economics,

KU Leuven, Naamsestraat 69, Leuven, B-3000, Belgium, [email protected], Erik Demeulemeester

This paper presents a decision support system for hospital case mix and capacity planning and describes the results of an application in a Belgian hospital. The case mix and capacity planning concerns both the long-term decision on patient volume and mix and the medium-term decision on resource allocation. The objective is to advance the trade-offs between efficiency and service. For this purpose, a model-based integrative methodology is developed on the basis of operational research techniques.

344

INFORMS Phoenix – 2012

4 - Optmization Approach to Dispatch EMS Vehicles

Farshad Majzoubi, University of Louisville, 775 Theodore Burnett

Ct Apt4, Louisville, KY, 40217, United States of America, [email protected], Lihui Bai, PhD., Sunderesh Heragu

The research considers the problem of dispatching ambulances when the demand for ambulance is high. Two scenarios are considered: 1- Dispatcher can determine the hospital, 2- Patients determine the hospital. For the first scenario, the mathematical model using general software optimizers is shown to be useful , but, for the second scenario, mathematical model is not solved in a reasonable time, hence, some heuristic methods are developed to be used in real situations.

5 - Resource Allocation Planning for High Quality Healthcare

Services in a Super-aged Society

Mihiro Sasaki, Prof., Nanzan University, 27 Seirei, Seto, 489-0863,

Japan, [email protected], Takamori Ukai, Masanori Hara,

Shigeaki Inoue

Due to coming super-aged society in Japan, the distribution of patients affected by acute and chronic disease has been changed drastically. Since the current healthcare system was designed in several decades ago without taking account of facing aged-society, many of medical staff are concerned about collapse of the healthcare system in the near future. We are trying to reallocate resources to improve the quality of healthcare services using operations research technique.

TD21

5 - A Stochastic Structural Reliability Model Explains Rotator Cuff

Repair Re-tears

Richard Hughes, Associate Professor, University of Michigan, 2003

BSRB, 109 Zina Pitcher Pl, Ann Arbor, MI, 48109, United States of

America, [email protected], James Carpenter, Drew Donnell,

Christopher Mendias, Jessica Seidelman, Bruce Miller

Re-tear of surgically repaired rotator cuff tears is common. We developed a twostate time-inhomogeneous Markov model to predict survival of repaired full-thickness rotator cuff tears. The probability of re-tear at each time step was modeled using a probabilistic structural reliability model. Model predictions were compared to the clinical results of our (J.E.C. and B.S.M.) patients. The model predicted 2-year survival within the 95% confidence interval of the Kaplan-Meier estimate (75.7%).

TD20

20- West 211 B- CC

Health Care Treatment and Therapy

Contributed Session

Chair: Richard Hughes, Associate Professor, University of Michigan,

2003 BSRB, 109 Zina Pitcher Pl, Ann Arbor, MI, 48109,

United States of America, [email protected]

1 - Administrative Barriers to Opening Clinical Trials

Diego Martinez, University of South Florida, 4202 E. Fowler

Avenue, ENB118, Tampa, FL, 33620, United States of America, [email protected], Kimberlea Hauser,

Benjamin Djulbegovic, Athanasios Tsalatsanis, Ali Yalcin

There are many challenges associated with opening a clinical trial in a university setting such as trial design, administrative procedures and approvals, patient and physician accrual. The proposed research explores and evaluates the administrative procedures required for opening a clinical trial at the University of

South Florida.

2 - Optimal Control for Predicting Drug Dosage in Superovulation

Stage of in Vitro Fertilization

Urmila Diwekar, President, Vishwamitra Research Institute, 368

56th Street, Clarendon Hills, IL, 60514, United States of America, [email protected], Kirti Yenkie

Superovulation is the most crucial stage in In Vitro Fertilization (IVF), since it involves external injection of hormones to stimulate development and maturation of multiple eggs. This paper presents a physics based model for superovulation validated with patient data and used to predict drug dosage based on optimal control theory.

3 - An Effective Methodology for Predicting the Severity of Prostate

Cancer Based on Biopsy Results

Arthur Swersey, Professor of Operations Research, Yale School of

Management, Box 208200, New Haven, Ct, 06520, United States of America, [email protected], Johannes Ledolter,

Rodney Parker

The effectiveness of PSA screening for prostate cancer is a hotly debated topic with all sides agreeing that overtreatment is a major problem. We develop a simulation model of the prostate biopsy procedure and use probability modeling to devise an effective decision rule for deciding whether to treat, based on the amount of cancer found on biopsy. This rule results in a high rate of detection and treatment of important tumors, while minimizing the treatment of insignificant ones.

4 - Treatment Selection Strategy for Long-Term Management of

Follicular Lymphoma

Larry White, Associate Professor, Eastern Illinois University,

School of Business, 600 Lincoln Avenue, Charleston, IL, 61920,

United States of America, [email protected]

Healthcare providers and patients have numerous options for treatment of follicular lymphoma, an incurable indolent cancer. We investigate factors affecting treatment selection including expected benefits to survival, quality of life, time to next treatment; expected costs including out-of-pocket expenses, third-party payments, time off work; provider capabilities; and expected research advancements. We derive appropriate treatment strategies for a variety of scenarios and objectives.

TD21

21- West 212 A- CC

Decomposition Approaches for Solving Network

Optimization Problems

Sponsor: Optimization/Networks

Sponsored Session

Chair: Jose L. Walteros, University of Florida, 303 Weil Hall,

Gainesville, FL, 32611-659, United States of America, [email protected]

1 - A Branch-and-Price Approach for Solving the Critical Clique

Detection Problem

Jose L. Walteros, University of Florida, 303 Weil Hall, Gainesville,

FL, 32611-659, United States of America, [email protected],

Panos Pardalos

The problem of detecting critical elements in a graph involves identifying a subset of elements whose deletion minimizes a connectivity measure of the induced subgraph. In this work we examine the problem of detecting critical cliques. We introduce two (0, 1)-linear formulations that are solved via branch-and-price. We present computational experiments comparing our results to competing formulations, as well as directions to modify our approach to tackle other critical element detection problems.

2 - Benders Decomposition for a Strategic Network Design

Problem under NAFTA Local Content Requirements

Katharina Mariel, Daimler AG, Mercedesstrafle 137, E 203,

Stuttgart, 70546, Germany, [email protected],

Stefan Minner

Free Trade Agreements strongly influence global production and sourcing strategies of multinational corporations. Based on local content requirements of the North American Free Trade Agreement (NAFTA) for automotive goods, a mixed integer, non-linear model is presented and solved by Benders

Decomposition. Numerical results illustrate the capability of the algorithm and give some insights into election to average rules.

3 - A Complimentary Column Generation Approach for a Complete

Graph Partitioning Problem

Salem Al-Yakoob, Assocsiate Prof, Kuwait University, Kuwait,

Kuwait, [email protected]

We will present two mathematical programming formulations and a complimentary column generation approach for the problem of partitioning a complete weighted graph into complete subgraphs having the same number of vertices with the objective of minimizing the total edge weights of the resulting subgraphs. Computational results will be presented and analyzed using a wide range of test problems.

4 - A Column Generation Framework for Routing Problems with

Time Windows

Leonardo Lozano, Instructor, Universidad de los Andes,

Departamento de Ingenieria Industrial, Carrera 1 Este # 19 A-40,

Bogota, Colombia, [email protected], Andrès L. Medaglia

Routing problems with time windows are well known NP-Hard problems. New variants emerge in various contexts (e.g., home care) posing new challenges and including different side constraints. We propose a column generation framework that can be embedded into exact approaches or matheuristics. The novelty lies in the specialized network algorithm proposed for the pricing problem. Our framework can be extended to several routing problems like the dial-a-ride or the

VRPTW with pick-up and delivery.

345

TD22

INFORMS Phoenix – 2012

TD22

22- West 212 B- CC

Branching in Mixed-Integer Programming II

Sponsor: Computing Society

Sponsored Session

Chair: John Chinneck, Carleton University, Systems and Computer

Engineering, Ottawa, On, K1S 5B6, Canada, [email protected]

1 - A Wide Branching Strategy for Branch-and-Price Algorithms

David Morrison, University of Illinois, Urbana-Champaign,

Siebel Center, Department of Computer Science,

201 N. Goodwin Ave., Urbana, IL, 61821, United States of

America, [email protected], Sheldon Jacobson, Jason Sauppe,

Edward Sewell

The usual branching strategy for branch-and-price algorithms forces a single variable to either 0 or 1 and attempts to preserve the pricing problem structure.

When this structure cannot be preserved, a wide branching strategy, which branches on all available variables and assigns them value 1, can be used. This strategy uses delayed branching techniques to limit the search tree width to a manageable size, and is competitive with the best-known solvers for graph coloring problems.

2 - Branching in SCIP

Gerald Gamrath, Research Assistant, Zuse-Institut Berlin,

Takustr 7, Berlin, 14199, Germany, [email protected]

Deciding how to branch is one of the most important decisions any branch-andbound based MIP solver has to take. In this talk, we discuss the so-called “hybrid branching” scheme, which combines the well-known concepts of strongbranching and pseudo-costs with more general criteria like conflicts and inferences originating from the SAT and CP context. Regarding a current example, we discuss drawbacks of current methods and present recent experiments concerning an extended strong branching.

3 - Using Lattice Basis Reduction to Compute General

Disjunctions in Practice

Sanjay Mehrotra, Professor, Northwestern University,

2145 Sheridan Rd., Evanston, IL, United States of America, [email protected], Kuo-Ling Huang

We will present results from a practical implementation of general disjunction algorithm using the Mehrotra and Li approach for solving general mixed integer convex programs. This method uses log-barrier ellipsoidal approximation for the continuous relaxation and performs Lenstra-Lenstra-Lovaz lattice basis reduction methods as a core to find a good branching direction. Computational results show significant difference in branch-and-bound nodes in many problems.

4 - Parallel Approach to Information-based Tree Search

Rodolfo Carvajal, Georgia Institute of Technology, 755 Ferst Drive

NW, Atlanta, GA, 30332, United States of America, [email protected], George Nemhauser, Shabbir Ahmed

To reduce the number of nodes searched in branch-and-bound algorithms,

KilinÁ-Karzan et al. (2009) proposed a branching scheme that collects information during the search and uses it in a restart phase. We implement variants of this approach that exploit parallelism in the collection or use of information and report computational results.

TD23

23- West 212 C- CC

Joint Session Analytics/CPMS: Reprise 2011

Innovation in Analytics Finalists and Award Winner

Sponsor: Analytics & CPMS, The Practice Section

Sponsored Session

Chair: Michael Gorman, University of Dayton, Dayton, OH,

United States of America, [email protected]

1 - Service Delivery Modeling and Optimization

Yixin Diao, IBM, Yorktown Heights, NY, United States of America, [email protected], Aliza R. Heching, David Northcutt

This project addresses the problem of identifying optimal staffing (staffing levels, shifts, and skills) for IBM’s globally located IT service delivery teams. Over the past three years, the project team has defined, developed, and deployed complex simulation models across multiple service line disciplines. These models have been deployed on a massive scale across the global scope of the IBM delivery organization. This has resulted in improved resource usage, delivery efficiency, and service quality over a large number of service delivery teams.

2 - Managing Immigration and Customs Enforcement’s Program

Operations with Innovative Analytics

Cenk Tunasar, Booz Allen Hamilton, McLean, VA, 22101,

United States of America, [email protected], Nicholas Nahas,

Patrick McCreesh, Jeffrey Munns, Govind Nagubandi

Since 2009, Booz Allen has supported Immigration and Customs Enforcement

(ICE) Law Enforcement Systems and Analysis (LESA) unit’s effort to increase criminal alien removals with descriptive, predictive, and prescriptive analytics.

These analytics are used to quantify resource needs; assess performance against objectives; and determine strategic direction. Five models-The Strategic Decision

Model (SDM), Operational Workforce Analysis, Criminal Alien Population Model,

Deployment Optimization Model, and the Network Optimization Mode- are used for this innovative analytics support. The innovation is in the integration of models, the out-come centered approach, and how these models deliver traditional industrial engineering solutions to creatively solve a public sector challenge of immigration enforcement. Over the last three years, Booz Allen has changed the way that ICE talks about, and thinks about, performance. Booz Allen has shaped the thinking of ICE’s senior leadership and supported the development of an agency-wide, quantitatively-driven strategy to identify and remove criminal aliens from the U.S. effectively and efficiently. The agency looks for opportunities to “optimize” rather than just seek incremental improvements.

The evolution in mindset is one strong impact of the analytics support at ICE.

TD24

24- West 213 A- CC

Stochastic Models for Medical Decision Making

Sponsor: Health Applications Society

Sponsored Session

Chair: Yuanhui Zhang, PhD Student, North Carolina State University,

375 Daniels Hall, Campus Box 7906, Raleigh, NC, 27695,

United States of America, [email protected]

1 - Iterative Characterization of Stable Equilibria for the Timing of

Paired Kidney Exchanges

Murat Kurt, Assistant Professor, University at Buffalo, SUNY,

University at Buffalo, North Campus, 415 Bell Hall, Buffalo, NY,

14260, United States of America, [email protected],

Utku Unver, Mark Roberts, Andrew Schaefer

Paired kidney exchanges (PKE), overcomes many difficulties in matching patients with incompatible donors. We consider a stochastic game formulation to model the transplant timing decisions in a two-way prearranged PKE. For large scale problems, we develop a convergent iterative algorithm to characterize a stable equilibrium of the game. We use clinical data to illustrate computational efficiency of the algorithm to equilibrium characterization through MILP and compare policy outcomes.

2 - Personalized Biopsy Referral Decision Modeling in the

Presence of Breast Cancer Regression

Fan Wang, University of Arkansas, Department of Industrial

Engineering, 4207 Bell Engineering Center, Fayetteville, AR,

72701, United States of America, [email protected],

Shengfan Zhang, Kathleen Diehl

Mammography, the current standard modality for breast cancer screening, is not perfect. Biopsy is often performed to verify the presence of cancer after positive results. We developed a discrete time finite-horizon Markov decision process model to optimize biopsy referral policies based on total expected quality-adjusted life years. The model incorporates individual’s breast cancer risk factors and mammogram screening history and allows the possibility of spontaneous disease regression.

3 - Modeling to Inform Proactive Management of Inpatient Care

Muge Capan, North Carolina State University,

400 Daniels Hall; College of Engineering, Raleigh, NC, 27695,

United States of America, [email protected], Julie S. Ivy

Improving inpatient care is an emerging focus in hospitalist systems. By considering patients’ frailty and vital signs, patient profiling provides a method for predicting deterioration in patient health. We propose to develop an optimization model based on sequential health data to create proactive health measures and support individualized rescue interventions.

4 - Coordination of the Influenza Vaccine Supply Chain with Effort

Xinghao Yan, Richard Ivey School of Business, 1151 Richmond St.

N, London, ON, Canada, [email protected], Greg Zaric

We study the influenza supply chain consisting of the health authority, the vaccine manufacturer, and the population. The health authority can make efforts to increase the vaccination rate from the population. We compare the decisions and the social welfare (number of infections) in the decentralized case and those in the centralized case. Finally, we investigate coordination contracts for the influenza supply chain with the health authority’s effort.

346

INFORMS Phoenix – 2012

TD25

25- West 213 B- CC

New Healthcare

Sponsor: Health Applications Society

Sponsored Session

Chair: Yichuan Ding, Stanford University, 14 Comstock Circle, Apt 10,

Stanford, CA, 94305, United States of America, [email protected]

1 - Efficiency of Subsidy Schemes in Reducing Waiting Times for

Public Health Services

Qu (Alex) Qian, Assistant Professor, Shanghai University of

Finance and Economics, 777 Guoding Road, School of

International Business Adm, Shanghai, 200433, China, [email protected], Pengfei Guo

This paper studies the efficiency of two typical subsidy schemes used in health care systems: unconditional and conditional. For each subsidy scheme we consider two information scenarios, no and full, that represent situations in which customers observe no realtime delay information and workload, respectively. We find that preferences for the two subsidy schemes depend on the information scenario and the size of the public fund.

2 - Improving Healthcare Delivery to Traveling Patients in the

Veterans Health Administration

Erkan Ceyhan, Massachusetts Institute of Technology,

77 Massachusetts Ave., Cambridge, MA, 02139,

United States of America, [email protected], Shahed Al-Haque

Performed in collaboration with the VHA, this research investigates the improvement of healthcare delivery to traveling patients. After analysis of the current state, the study aims to provide operations management strategies, through simulation of potential policy interventions, that enable more effective and efficient resource allocation to meet traveling patients’ needs.

3 - Controlling Healthcare Queues

Sara Nourazari, Northeastern University, 334 Snell Eng. Center,

360 Huntington Ave., Boston, MA, 02115, United States of

America, [email protected], James Benneyan, Rifat Siphai

Queuing problems abound in almost all service systems. Typical analysis or design approaches are to statically determine service capacity to achieve acceptable mean performance levels, such as mean number or time in queue, resulting in periods of poor service spikes and unattractive utilization-service-cost tradeoffs. We explore the use of control theory to adapt service capacities in healthcare systems to achieve desired mean and threshold performance levels under a variety of assumptions.

4 - Modeling Patient Follow-ups in a Primary Care Clinic

Yichuan Ding, Stanford University, 14 Comstock Circle, Apt 10,

Stanford, CA, 94305, United States of America, [email protected], Chinghua Chen, Mayank Sharma

In an outpatient setting, patients arriving for a single appointment may unexpectedly find that they must return for one or more follow-up appointments.

This paper develops a Markov decision process (MDP) to model this phenomenon when the regular patients have to book their visits in advance. We derive an approximately optimal policy to this MDP and evaluate its performance by simulations.

TD26

26- North 221 A- CC

Customer Queuing Behaviors and Their

Implications for M&SOM

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Mirko Kremer, Pennsylvania State University, State College, PA,

United States of America, [email protected]

1 - Expert Queues: Information Pricing and Congestion

Senthil Veeraraghavan, Associate Professor, The Wharton School,

3730 Walnut Street Suite 550, Philadelphia, PA, 19104,

United States of America, [email protected],

Mehmet Fazil Pac

Consumers cannot identify their problems often (for e.g. car repair). Therefore they rely on experts, who also provide the service, for the diagnosis of their problems. The information asymmetry arising upon diagnosis leads to inefficiencies in service provision. The expert has an incentive to over-provide or to ration services, based on the demand, capacity and the consumer waiting costs.

We investigate diagnosis and pricing decisions in expert service markets using a queuing framework.

TD27

2 - Visible Queues: Customer Response to Waiting at Gas Stations

Margaret Pierson, Assistant Professor, Tuck at Dartmouth,

Hanover, NH, United States of America, [email protected]

Empirical study of consumer choice in gasoline market using pricing data and queue inference from observations of service time. Two year panel data study, with observations of all transactions, at 15 retail outlets in Massachusetts. Queue inference is augmented by study of video footage from the gasoline retailer.

Anecdotal observations of customer behavior observed in the footage will also be presented.

3 - Service Systems with Anecdotal Reasoning Consumers

Ying-Ju Chen, UC Berkeley, Berkeley, United States of America, [email protected], Yimin Yu, Tingliang Huang

We study service systems where consumers have boundedly rational expectations. In particular, we assume that consumers lack full capability or ample opportunity to perfectly infer the waiting time, and thus can only rely on past experiences and anecdotal reasoning to make their joining or balking decisions.

4 - Learning Quality from Queues: A Laboratory Experiment

Laurens Debo, University of Chicago, Chicago, IL, United States of

America, [email protected], Mirko Kremer

In this talk, I will report the results of a laboratory experiments in which human subjects need to decide whether to join the queue or not, when the quality of the service is unknown to the subject, but known to a small fraction of the population.

TD27

27- North 221 B- CC

Topics in Service Management

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Ramandeep Randhawa, University of Southern California,

Los Angeles, United States of America, [email protected]

1 - Vendor Selection, Contract Efficiency and Performance

Measurement in Service Outsourcing

Sameer Hasija, INSEAD, Fontainbleau, France, [email protected], Zhijian Cui

This study compares the efficacy of some commonly observed vendor selection and contracting mechanisms with respect to two key challenges in service outsourcing: vendor selection and contract efficiency. We show that competitive bidding yields good selection but contract inefficiency. In contrast, if the client establishes the contract terms then the “menu” it designs yields contract efficiency but poor selection.

2 - Contracting and the Dynamics of Collaboration in

Service Processes

Morvarid Rahmani, PhD Candidate, UCLA, 110 Westwood Plaza,

Los Angeles, 90095, United States of America, [email protected], Guillaume Roels,

Uday Karmarkar

Collaboration governs many work processes; for example, it is common in business-to-business services. In this paper, we study when collaboration takes place in a joint project, and how the dynamics of collaboration are aected by the project deadline, the veriability of efforts, and the type of contract adopted by the parties involved in the project.

3 - Groupon for Services: Is it a Good Deal?

Guangwen Kong, University of Southern California, Marshall

School of Business, Los Angeles, United States of America, [email protected], Ramandeep Randhawa

Online social advertising tools such as Groupon generate new business for service providers and at the same time generate new challenges. Discount-seeking customers may impose externalities on the system that could drive away regular customers. We analyze these tradeoffs and devise recommendations as to when would Groupon promotions be beneficial.

4 - Setting Quality and Speed in Service Industry with

Repeated Customers

Azin Farzan, Student, University of Washington, Box 353226,

Foster Business School, Univ. of Wa., Seattle, WA, 98195,

United States of America, [email protected], Yong-Pin Zhou

We analyze a system in which customers purchase services repeatedly. Customers benefit from a higher quality level and incur cost due to waiting. Customers choose to join the service based on the expected value of the service. After each encounter, they repeat the purchase depending on the quality perceived. The firm invests in capacity and quality levels jointly. We analyze these decisions and their relationship. We also consider the duopoly scenario and analyze the equilibria in each setting.

347

TD28

INFORMS Phoenix – 2012

TD28

28- North 221 C- CC

Sustainability in Operations Management

Sponsor: Manufacturing & Service Oper Mgmt

Sponsored Session

Chair: Serguei Netessine, Professor, INSEAD, Boulevard de Constance,

Fountainbleau, 77305, France, [email protected]

1 - Dynamics of Sustainable Business Relationship

Pavel Izhutov, PhD Student, Stanford University, 655 Knight Way,

Stanford Graduate School of Business, Stanford, CA, 94305,

United States of America, [email protected], Hau L. Lee

A retailer is interested in designing a contract that ensures the use of responsible production process by the supplier. The key features of the examined contract are its duration and continuation/termination conditions. The optimal contract ensuring responsible supplier behavior is derived.

2 - Electric Vehicles with a Battery Switching Station: Adoption and

Environmental Impact

Buket Avci, INSEAD, Boulevard de Constance, Fontainebleau,

77300, France, [email protected], Serguei Netessine,

Karan Girotra

We analyze a novel switching-station based business model for the deployment of electric vehicles. We develop a stylized analytical model that captures the salient features of EV adoption decision by modeling range anxiety and the impact of different ownership structures (selling miles vs. selling batteries). We find that electric vehicles with switching stations can incent adoption and reduce oil dependence but, paradoxically, this increased adoption may not necessarily benefit the environment.

3 - Delivering the Sustainable Way

Gerard Cachon, University of Pennsylvania,

3730 Walnut Street, Philadelphia, United States of America, [email protected]

People need to have goods brought to their homes. They can do this themselves drive to a store, purchase a basket of items and then return back home. Or, they can order on-line and have the goods delivered to their doorstep. With a view to sustainability, which of those approaches is better?

4 - Strategic Investment in Renewable Energy Sources

Sam Aflaki, HEC, France, [email protected], Serguei Netessine

We model the tradeoff between investing in an intermittent renewable technology (such as wind) and a reliable non-renewable technology (such as natural gas). Motivated by existing electricity markets, we analyze the effectiveness of carbon pricing strategies to promote renewables in several interrelated contexts including: (a) vertically integrated supplier, (b) market competition and (c) long-term fixed-price contracts for renewable electricity.

TD29

29- North 222 A- CC

Joint Session MSOM/SPPSN: Emergency Medical

Services (EMS) Operations

Sponsor: Manufacturing & Service Oper Mgmt/ Healthcare

Operations/SIG & Public Programs, Service and Needs

Sponsored Session

Chair: Alex Mills, Indiana University Kelley School of Business,

1309 E. Tenth Street, Bloomington, IN, 47405,

United States of America, [email protected]

1 - Scheduling and Routing Ambulances that Provide Inter-facility

Patient Transfers

Laleh Haerien, University of Alberta, School of Business,

Edmonton, AB, T6G 2R6, Canada, [email protected],

Dan Haight, Armann Ingolfsson, Angela Kercher,

Mohamed Salama, Matt Stanton

In Edmonton and Calgary, Canada, a specialized ambulance fleet transfers patients between hospitals. We use a VRP heuristic to schedule transfers that are known a day in advance, and a similar heuristic to accommodate emergent transfers in real time. Simulation experiments comparing the performance of this approach to historical schedules suggests that improvements on all performance metrics that we investigated are possible, ranging from 60% (for lateness) to 6%

(for deadhead travel).

2 - Optimizing the Deployment of Public Access Defibrillators

Derya Demirtas, University of Toronto, 5 King’s College Rd,

Toronto, ON, M5S 3G8, Canada, [email protected],

Roy Kwon, Timothy Chan

Sudden cardiac arrest is a major public health problem. The treatment for cardiac arrest is very time-sensitive. Deploying automated external defibrillators (AEDs) in public areas for bystander use reduces the time to defibrillation and improves survival rates. In this talk, we present optimization models to guide the deployment of AEDs in public settings. Computational results using real cardiac arrest and AED data from Toronto, Canada demonstrate the effectiveness of the proposed models.

3 - A Model for Optimally Transporting Emergency Medical Service

Patients to Hospitals

Benjamin Grannan, Virginia Commonwealth University,

P.O. Box 843083, Richmond, VA, 23284, United States of America, [email protected], Laura McLay

We present a Markov decision process model for determining how to optimally dispatch ambulances to patients and then transport patients to distinguishable receiving locations (hospitals). Through this model, we investigate how the revelation of information about a patient over time affects the interrelated dispatch and transport decisions. Particular attention is paid on the impact of initial customer classification errors. The optimal policy is compared to two myopic policies.

4 - Resource Allocation in Mass Casualty Incidents

Alex Mills, Indiana University Kelley School of Business, 1309 E.

Tenth Street, Bloomington, IN, 47405, United States of America, [email protected], Serhan Ziya, Nilay Argon

When multiple mass-casualty incidents compete for limited resources, such as ambulances, emergency planners must decide both how to allocate resources and how to prioritize different classes of patients. Widely used triage protocols, such as

START, do not describe how to handle these decisions. We develop centralized and decentralized prioritization policies that can be effective in such a scenario.

TD30

30- North 222 B- CC

Facilities Planning and Design

Contributed Session

Chair: Venkat Rohit Kota, PhD. Candidate, Grado Department of

Industrial and Systems Engineering (0118) Virginia Polytechnic

Institute and State University, 250 Durham Hall, Blacksburg, VA,

24060, United States of America, [email protected]

1 - Facility Layout: Optimization and Simulation Approach

Phong Ho, Professor, HCM City International University, KP6,

Linh Trung, Thu Duc, Ho Chi Minh City, Vietnam, [email protected], Huy Truong, Quoc Ho

This research integrates multi-objective optimization and simulation to solve a factory layout problem where a new objective, reachability, has been added.

Reachability is defined as the ability of one facility to reach an alternative facility when the default facility is down. Genetic Algorithms is integrated in simulation to evaluate solutions. A numerical example is presented to demonstrate the effectiveness of the proposed solution.

2 - Solving a Facility Layout Problem for a Manufacturer Applying a

Meta-Heuristic and Lean Objectives

Ahmed Abdulaal, M.S. Candidate, University of New Haven, 300

Boston Post Rd, West Haven, CT, 06516, United States of America, [email protected], Amy Thompson

This research presents a facility layout case study that finds an improved facility layout for a manufacturer based upon a set of lean objectives that include safety, quality, flow, and efficiency. Through development of a new meta-heuristic, a significantly better solution was found demonstrating a real-world application to a large-scale problem. The model allocated 43 machines to a 23,000 square foot production floor in minutes, reduced flow cost by 40%, and freed significant floor space.

3 - Analytical and Heuristic Based Results for Puzzle-Based

Warehouse Systems

Venkat Rohit Kota, PhD. Candidate, Grado Department of

Industrial and Systems Engineering (0118) Virginia Polytechnic

Institute and State University, 250 Durham Hall, Blacksburg, VA,

24060, United States of America, [email protected],

Gaylon Don Taylor

One manifestation of very high density storage systems is the ‘puzzle-based’ system, in which unit loads move through the system by changing the location of empty escorts to retrieve demanded items. In this research, the authors investigate retrieval times for scenarios when the number of empty locations are two or more. Closed form solutions, and heuristic based approaches are proposed to aid retrieval strategies.

348

INFORMS Phoenix – 2012

4 - The Model Development of Plant Productivity Improvement in

Steel Factory

Mohammad A. Quasem, Department of Information & Decision

Sciences, Washington, DC, United States of America, [email protected], M.H. Shwehdi

The Reliable distribution of electric power has been plagued by common problems such as interruption, voltage sags, and swells, as well as voltage flicker, and transients. The impact of these problems has gotten worse in recent years due to the nature of the loads that are served by the power system. This paper will present the different power quality problems encountered by some of the plants.

The study suggests and recommends solutions to increase savings and reduce power quality problems and therefore plant productivity.

TD31

31- North 222 C- CC

Empirical Research in Service Operations

Sponsor: Manufacturing & Service Oper Mgmt/Service Operations

Sponsored Session

Chair: Marcelo Olivares, Columbia Business School, 3022 Broadway,

Uris Hall 417, New York, NY, 10027, United States of America, [email protected]

1 - Learning from Customers: Individual and Organizational Effects

Bradley Staats, Assistant Professor, University of North Carolina-

Chapel Hill, McColl Building, CB 3490, Chapel Hill, NC, 27599,

United States of America, [email protected]

We explore the customer dimension of volume-based learning in a radiological setting, where individual doctors at an outsourcing firm complete radiological reads for hospital customers. We examine more than 2.7 million cases for 1,431 customers read by 97 radiologists and find evidence supporting the benefit of accumulating customer-specific experience. We discuss the implications of our results for the study of learning as well as for the providers and consumers of outsourced services.

2 - Reputation in Online Service Marketplaces

Antonio Moreno-Garcia, Kellogg School of Management, 2001

Sheridan Road, Evanston, 60208, United States of America, [email protected], Christian Terwiesch

Online service marketplaces allow service buyers to post their projects and service providers to bid for them. In order to reduce the transactional risks, previous seller performance is usually reflected in a numerical reputation score. By analyzing a detailed dataset with more than 1,800,000 bids corresponding to

270,000 projects posted between 2001 and 2010 in a leading on-line intermediary for software development services, we empirically study the effects of reputation on market outcomes.

3 - Data Analysis of Service Times in Call Centers

Noah Gans, University of Pennsylvania, The Wharton School,

3730 Walnut Street, Suite 500, Philadelphia, PA, 19104,

United States of America, [email protected], Han Ye,

Haipeng Shen

We have been analyzing large datasets from telephone call centers to better characterize pervasive but little-studied phenomena that affect their operations.

In this talk, we’ll discuss agent call times, which show considerable heterogeneity and evolve according to somewhat predictable patterns.

4 - Empirical Study on Agent Productivity in IT Service

Yina Lu, Columbia Business School, 3022 Broadway, 423 Uris Hall,

New York, NY, 10027, [email protected], Aliza R. Heching,

Marcelo Olivares

We analyze a new dataset from a leading IT company that captures how agents spend their time and switch among tasks. We find that agent productivity is a time-dependent metric as opposed to be static and intrinsic to an agent. We thus propose the use of hazard rate to measure productivity and explore the driving factors that affect it, including interruptions, unfinished workload and cumulative workload during a shift. Our findings can help to develop more intelligent task routing scheme.

TD32

32- North 223- CC

Supply Chain Risk and Information

Sponsor: Manufacturing & Service Oper Mgmt/Supply Chain

Sponsored Session

Chair: Zhibin (Ben) Yang, Assistant Professor, University of Oregon,

1208 University of Oregon, Eugene, 97403, United States of America, [email protected]

1 - Information Asymmetry with the Presence of Spot Market

Xuan Zhao, Associate Professor, Wilfrid Laurier University, 75

University Ave.W., Waterloo, ON, N2J4Y8, Canada, [email protected],

Wei Xing

This paper studies the effect of information asymmetries (in both demand and the spot price) on the members of a two-stage supply chain in the presence of spot market. Using a novel method to obtain the posterior bivariate normal distributions, we model the information updating Stackelberg game and derive the unique equilibrium strategies. We then explore when the supplier intends to use a forward contract, whether a better demand/spot price forecast always preferred.

2 - Are Responsive Pricing and Supply Diversification Substitutes in Hedging Supply Uncertainty?

Tao Li, UT Dallas, 800 W Campbell Road, Richardson, TX, 75080,

United States of America, [email protected], Suresh Sethi,

Jun Zhang

Responsive pricing and supply diversification are two strategies that firms adopted to mitigate supply uncertainty. It seems intuitive that they are substitutes in hedging supply uncertainty. We find that they are actually complements under certain conditions. Specifically, when the cost of lost goodwill is low and suppliers’ costs are medium, the firm’s diversification propensity is higher and it can gain higher value of diversification with responsive pricing than without.

3 - Coordination in a Multi-retailer Distribution System:

Supplier-facilitated Transshipments

Rong Li, SMU, 50 Stamford Road, Singapore, Singapore, [email protected], Zhi Zeng, Jennifer Ryan

We study a two-period production and inventory distribution problem in which transshipments, reactive and proactive, among independent retailers occur in the second period. We propose and examine a new coordination scheme: supplierfacilitated transshipments, implemented via a bi-directional adjustment contract, offered by the supplier to each retailer. We show that coordination can be achieved; a unique state-dependent adjustment (or transshipment) price must be used for coordination.

4 - Timing and Signaling Considerations for Recovery from Supply

Chain Disruption

Zhibin (Ben) Yang, Assistant Professor, University of Oregon, 1208

University of Oregon, Eugene, 97403, United States of America, [email protected], Nagesh Murthy

We study a supplier’s timing decisions for recovery from a disruption, when the buyer trades off backup option and waiting for supply recovery. The supplier quotes a due date for recovery, and can accelerate supply recovery by increasing effort. We find the supplier strategically quotes an early due date for retaining the buyer, leading to Pareto-improvement of the channel. When the supplier is privately informed about disruption severity, the due date serves as a signal of disruption severity.

TD33

33- North 224 A- CC

Inventory Allocation and Pooling

Contributed Session

TD33

Chair: S Viswanathan, Professor, Nanyang Technological University,

College of Business, Nanyang Avenue, Singapore, 639798, Singapore, [email protected]

1 - A Continuous Review Inventory System with Two Types of Customers

Sapna Isotupa, Associate Professor, Wilfrid Laurier University,

School of Business and Economics, Wilfrid Laurier University,

Waterloo, ON, N2L 3C5, Canada, [email protected]

In this paper a lost sales (S-1, S) inventory system with two types of customers – high priority and low priority is analyzed. The central objective is to determine at what inventory levels should low priority customers be turned away to prevent stock outs for the high priority customers.

349

TD34

2 - Inventory Allocation Methods and Service Level Agreements

Chun-Miin Chen, Pennsylvania State University, 430 Business

Building, University Park, PA, 16802, United States of America, [email protected], Doug Thomas

Firms may employ a variety of methods to allocate scarce supply to meet multiple customer demands. We examine the impact to customer service levels of different allocation methods both analytically and computationally. We find substantial differences in customer service performance across different allocation rules.

3 - Cost Reductions by Pooling of Inventory

Alexandra van Wijk, Eindhoven University of Technology, P.O.

Box 513, Eindhoven, 5600 MB, Netherlands,

[email protected], Ivo Adan, Geert-Jan van Houtum

Downtime of machines is an expensive cost factor for many companies. For this reason, spare parts are kept on stock, to be able to quickly respond to breakdowns. We study spare parts inventory models, allowing for stock transfers between the local warehouses, so-called lateral transshipments. In this way, we create pooling of inventory. We investigate when lateral transshipments lead to cost reductions, how these can be optimally applied, and when simple, easy to implement policy are optimal.

4 - Advance Demand Information and Early Fulfillment in Inventory

Systems with Two Demand Classes

Sourish Sarkar, Assistant Professor, Black School of Business, Penn

State Erie, The Behrend College, 4701 College Drive, Erie, PA,

16563, United States of America, [email protected], John Shewchuk

We consider production-inventory systems serving two demand classes. One class provides advance demand information, and the other class does not. We assume early fulfillment is acceptable, and propose a policy for production replenishment and order fulfillment, which is shown to outperform the existing policies. Some managerial insights are also discussed.

5 - A New Class of Two Bin Policy for Inventory Rationing in a

Continuous Review System

S Viswanathan, Professor, Nanyang Technological University,

College of Business, Nanyang Avenue, Singapore, 639798,

Singapore, [email protected], Sugoutam Ghosh, Rajesh Piplani

We propose a new class of two-bin policy for the continuous review, single item inventory system with two demand classes. The proposed two-bin policy assigns separate bins of inventory for the two classes. However, demand from the higher class can still be fulfilled with the inventory from the other bin, when its own bin is empty. The proposed policy can outperform the critical-level rationing policy when there is a minimum fill rate requirement for the lower demand class.

TD34

34- North 224 B- CC

Project Management: Teams, Dynamics and Incentives

Cluster: New Product Development

Invited Session

Chair: Karthik Ramachandran, Georgia Institute of Technology,

Atlanta, GA, United States of America, karthik.rama[email protected]

1 - Projects and Team Dynamics

George Georgiadis, University of California-Los Angeles, Anderson

School of Management, 110 Westwood Plaza, Los Angeles, CA,

United States of America, [email protected]

This paper studies the dynamic collaboration of a team on a project that stochastically evolves to completion over time. The main result is that members of a larger team work harder both individually and on aggregate if and only if the project is sufficiently far from completion. By using this framework, I examine how a manager who recruits agents into a team to complete a project on her behalf should choose the team composition, as well as the agents’ compensations.

2 - An Incentive Scheme to Resolve Parkinson’s Law in

Project Management

Bo Chen, Professor, University of Warwick, Warwick Business

School, Gibbet Hill Road, Coventry, CV4 7AL, United Kingdom,

[email protected], Nicholas Hall, Xiaotie Deng

A major challenge in executing many projects is the widespread prevalence of

Parkinson’s Law, which results in wasting the benefit of potential early project completion. We design mechanisms that incentivize task owners to complete their tasks early when possible and to report their ability to do so when they realize this is possible with progress on their tasks. Our mechanisms further incentivize task owners to get ready for starting their tasks early.

INFORMS Phoenix – 2012

3 - Consequences and Control of Procrastination in Projects

Yaozhong Wu, National University of Singapore Business School,

15 Kent Ridge Dr., Singapore, Singapore, [email protected], Vish Krishnan, Karthik Ramachandran

We analyze how managers can design project tasks to influence procrastinating workers. We also report findings from an experiment on the effects of different task designs on individual worker’s behavior and on the quality of their performance.

4 - Stakeholder Commitment: Product Development With

Uncertainty

Jeremy Kovach, Georgia Institute of Technology, 800 West

Peachtree St. NW, Atlanta, GA, 30308, United States of America,

[email protected], Stylianos Kavadias

We explore incentive structures used by firms to motivate functional stakeholders’ resource commitment decisions during NPD projects. We consider projects where as the project progresses, the locus of responsibility shifts from one functional stakeholder to another. We design optimal contracts for the functional stakeholders, factoring in the stakeholder’s preference to resource commitment, in order to maximize project value based on the project uncertainty and project monitoring costs.

TD35

35- North 225 A- CC

Pricing and Revenue Management: Dynamic Pricing

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: Hongmin Li, Arizona State University, W.P. Carey School of

Business, Tempe, AZ, 85287, United States of America,

[email protected]

1 - A Simple Heuristic for Joint Inventory and Pricing Problems with Lead Time

Yang Li, PhD Student, Duke University, 100 Fuqua Drive, Durham,

NC, 27708, United States of America, [email protected],

Kevin Shang, Fernando Bernstein

We study a joint inventory and pricing problem in a single-stage system with positive lead time. This problem is, in general, intractable due to its computational complexity. We develop a simple heuristic that resolves this issue. The heuristic first generates a pricing policy which depends on the initial inventory level. We then transform the joint problem into a standard inventory problem. This heuristic enables us to explore the impact of lead time on the joint decision.

2 - Price Optimization under Generalized Extreme Value Models

Amr Farahat, Assistant Professor, Washington University in

St. Louis, Olin Business School, St. Louis, MO, United States of

America, [email protected], Renyu (Philip) Zhang

We investigate the multi-product price optimization problem under the

Generalized Extreme Value discrete choice model proposed by McFadden (1978).

We present structural results leading to a tractable approach to solving this problem which exhibits the non-concavity issues inherent in its special case: the

(nested) multinomial logit model.

3 - Cost-per-click Pricing for Display Advertising

Sami Najafi-Asadolahi, Rotman School of Management, University of Toronto, Toronto, Canada, [email protected],

Kristin Fridgeirsdottir

We determine the optimal cost-per-click (CPC) price of display ads for a web publisher facing uncertain demand from advertisers requesting space on its website and uncertain supply of clicks from viewers. We formulate the problem as a queuing system. We show that the frequently used heuristic applied by most web publishers to convert between the CPC and CPM pricing schemes using the so-called click-through-rate can be misleading and may incur web publishers substantial revenue loss.

4 - Optimal Pricing for a Short Life-cycle Product When Customer

Price-sensitivity Varies Over Time

Hongmin Li, Arizona State University, W.P. Carey School of

Business, Tempe, AZ, 85287, United States of America,

[email protected], Tim Huh

Demand for technology products often resembles a diffusion process, with weak demand in the beginning and end of the life cycle and high demand intensity in between. The customer price-sensitivity also changes over the life cycle of the product. We study the pre-announced pricing decision for a product that exhibits such demand characteristics and determine the optimal set of discrete prices and the times to switch from one price to another, when a limited number of price changes are allowed.

350

INFORMS Phoenix – 2012

TD36

36- North 225 B- CC

Advances in Discrete Consumer Choice Model

Sponsor: Revenue Management & Pricing

Sponsored Session

Chair: Ruxian Wang, Research Scientist, Hewlett-Packard Labs,

1501 Page Mill Road, Palo Alto, CA, 94304, United States of America, [email protected]

1 - Alternative Model for Discrete Sales: A Comparison with

Multi-choice Logit Model

Varun Gupta, PhD Student, The University of Texas at Dallas, 800

West Campbell Road SM30, Richardson, TX, 75080-3021, United

States of America, [email protected], Metin Cakanyildirim

In this study we propose a discrete choice demand model that can be used to model sales for competitive goods. We use real life data to compare the performance of the proposed model with the well established multi-choice logit model. We chose real-life data on different products such as yogurt, beef, ketchup and candy melts.

2 - The Multi-product Newsvendor Problem with Consumer Choice

Joonkyum Lee, Doctoral Student, Cornell University,

301A Sage Hall, Ithaca, NY, 14853, United States of America, [email protected], Amr Farahat

We address the muti-product under a general consumer choice behavior, e.g., substitution and choice overload. We present a tractable relaxation of an exact LP formulation that yields a feasible solution coupled with an upper bound on the optimal value. The relaxation is based on a fluid approximation and dynamic display strategy. It can be solved by identifying efficient inventory levels. The method is shown to perform well compared to existing methods and provides tighter optimality gap bounds.

3 - Discrete Choice Models for Dynamic Decisions

So Yeon Chun, Assistant Professor, Georgetown University,

McDonough School of Business, 37th and O Streets, Washington

DC, 20057, United States of America, [email protected],

Anton Kleywegt

Traditional discrete choice models assume that decision makers choose one alternative at a point in time, without the option to wait, collect more data, and revisit the decision later. We formulate a discrete choice model in which, at each decision time, decision makers have the option to postpone a decision until later.

Alternatives and attributes may change over time, and decision makers attempt to forecast attribute values at the next decision time.

4 - Product Portfolio Selection and Bounded Price Optimization

Shiqian Ma, University of Minnesota, Church Street 207,

Minneapolis, MN, 55455, United States of America, [email protected], Ruxian Wang

We study product selection and price optimization with constraints on portfolio size and price bounds under the multinormial logit model. We propose a new efficient algorithm and investigate its performance in the static optimization as well as the dynamic problem.

5 - Multi-product Revenue Management and Price Optimization under Capacitated Discrete Choice Models

Ruxian Wang, Research Scientist, Hewlett-Packard Labs, 1501

Page Mill Road, Palo Alto, CA, 94304, United States of America, [email protected]

We consider a dynamic assortment and price optimization under the capacitated multinomial logit model with product-differentiated price sensitivities. The multiproduct assortment and price optimization is transferred to a single-dimensional problem with a control of the aggregate resource consumption rate. By using the generalized efficient frontier with price optimization, we are able to establish a time-threshold structure for the dynamic problem.

TD37

37- North 226 A- CC

Location Objectives I

Sponsor: Location Analysis

Sponsored Session

Chair: Zvi Drezner, Professor, California State University,

Department of ISDS, Fullerton, CA, 92834, United States of America, [email protected]

1 - The Wisdom of Voters: Evaluating the Weber Objective at the

Condorcet Solution

Mozart Menezes, Associate Professor of Operations Management,

University of Calgary, Haskayne School of Business, Calgary,

Canada, [email protected], Zvi Drezner

TD38

We investigate the quality of Condorcet solutions comparing it to that of a central decision maker in the context of a locational problem, concluding, through exhaustive experiments, that if each population member votes according to his/her own self-interest, then the Weber objective at the Condorcet point is very close to the optimal Weber value. Being short of, or reducing the set of, candidate solutions has little impact on the quality of the Condorcet solution.

2 - Upper Bound on the Inefficiency of Condorcet Solution using

Weber Optimal Value as Benchmark

Rongbing Huang, York University, 4700 Keele Street, Toronto,

Canada, [email protected], Mozart Menezes

This paper focuses on the worst-case quality of the Condorcet solution measured through the upper bound of the ratio of the Weber objective function value at the

Condorcet solution to the Weber optimal objective function value. Prior work defined on networks showed the ratio to be 3. We show that on plane the ratio is smaller than 1.414. The result suggests that, when reducing the impact of topology, the solution originated via voting, it is much closer to optimality than previously suggested.

3 - Maximal Accessibility Network Design in the Public Sector

Robert Aboolian, California State San Marcos,

2771 Palmetto Drive, Carlsbad, CA, United States of America, [email protected], Oded Berman, Vedat Verter

This paper is on the maximal accessibility facility network design problem for the public sector. The governments’ mandate is to maximize the societal benefit by acting as agents of the public. Participants often patronize a publicly funded facility with highest accessibility. We assume that the time spent for receiving the service from a facility is a good proxy for its accessibility. We formalize the basic network design problem for the public sector and develop an exact approach to solve it.

4 - Stochastic Analysis of Ordered Median Problems

Zvi Drezner, Professor, California State University, Department of

ISDS, Fullerton, CA, 92834, United States of America, [email protected], Stefan Nickel, Hans-Peter Ziegler

Many location problems can be expressed as ordered median objective. In this paper we investigate the ordered median objective when the demand points are generated in a circle. We find the mean and variance of the k’th distance from the center of the circle and the correlation matrix between all pairs of ordered distances. By applying these values, we calculate the mean and variance of any ordered median objective and the correlation coefficient between two ordered median objectives.

TD38

38- North 226 B- CC

Service Operations Management: Forecasting,

Capacity Planning, Scheduling and Performance

Optimization

Sponsor: Service Science

Sponsored Session

Chair: Turgut Aykin, President and CEO, ac2 Solutions, Inc.,

12 Crown Plaza, Suite 207, Hazlet, NJ, 07730,

United States of America, [email protected]

1 - Call Center Forecasting: State of the Art and Current

Challenges

Turgut Aykin, President and CEO, ac2 Solutions, Inc.,

12 Crown Plaza, Suite 207, Hazlet, NJ, 07730,

United States of America, [email protected]

Over the past two decades, call centers showed a tremendous growth and have become a major channel for customer service and sale. This presentation focuses on the state of the art in forecasting with i) multiple seasonality, and ii) short interval (e.g. 15-, 30-minute) forecasting. We will also review a number of simple methods used at call centers.

2 - Staffing and Capacity Planning at Service Systems

John McKenna, VP, Product Development, ac2 Solutions, Inc.,

12 Crown Plaza, Suite 207, Hazlet, NJ, 07730,

United States of America, [email protected]

Resource planning at service systems is a collaborative effort involving operations, finance and line of business. In this session, we will discuss queuing models used for determining required staffing levels, and how they are used to support collaborative capacity planning.

351

TD39

3 - Concurrent Optimal Scheduling in Skills-based and Non-skills

Based Systems

Liz Turner, ac2 Solutions, Inc., 12 Crown Plaza, Suite 207, Hazlet,

NJ, 07730, United States of America, [email protected],

Turgut Aykin

It is known that staffing costs account for 60-70% of the operating budget at a typical service center. Given this, considerable emphasis was put on optimal shift scheduling problems in the literature. This talk focuses on scheduling in skills and non-skills based systems and discusses Integer Programming based approaches for these problems.

4 - Performance Management and Optimization in

Service Operations

John McKenna, VP, Product Development, ac2 Solutions, Inc.,

12 Crown Plaza, Suite 207, Hazlet, NJ, 07730, United States of

America, [email protected], Turgut Aykin

Forecasting, Capacity Planning and Scheduling are planning processes that are based on forecasts and other assumptions regarding the future. Deviations from assumptions during operations are unavoidable. This presentation describes several methods for managing performance in “real-time” including performance, skills and routing optimization, adherence and activity management.

TD39

39- North 226 C- CC

Topics in Operations Management II

Contributed Session

INFORMS Phoenix – 2012

Chair: Wenjun Gu, Assistant Professor, Southern New Hampshire

University, School of Business, 2500 North River Road, Webster Hall

217, Manchester, NH, 03106, United States of America, [email protected]

1 - Competing for Contracts on Price and Quality Variables

Edward Anderson, Professor Decision Sciences, University of

Sydney Business School, Sydney, NSW2006, Australia, [email protected], Cheng Qian

A buying firm chooses between several competing suppliers, who submit bids having both price and quality variables. We use a discrete choice model for the buyer to represent the uncertainty that exists from the supplier’s perspective. We show that each supplier chooses its quality variables in a way that is independent of the competition; in a sense the suppliers “compete only on price”. We also look at a model where the suppliers are uncertain about the buyer’s weighting on quality variables.

2 - Joint Determination of Shipping Fees and Product Prices for

Online Transactions

Nabita Penmetsa, University of Pittsburgh, Mervis Hall,

Pittsburgh, PA, United States of America,

[email protected], Jennifer Shang

While the convenience and time saving offered by online transactions are attracting customers towards E-commerce, shipping fees is a major factor limiting some customers from buying products online. E-tailers offer discounts on shipping fees to attract customers, but the discounts impact profitability of firms as shipping costs are substantial. We develop a mixed integer programming model to find optimal product prices and shipping fees that maximize E-tailer profits.

3 - Product Quality Design in Reverse Supply Chain

Wenjun Gu, Assistant Professor, Southern New Hampshire

University, School of Business, 2500 North River Road, Webster

Hall 217, Manchester, NH, 03106, United States of America, [email protected], Dilip Chhajed, Nicholas Petruzzi

This project studies how a manufacturer designs its product quality in a reverse supply chain when used products are collected, remanufactured and sold into the same market as the new ones. Results show that product quality generally increases in collection rate, but as does not hold in the presence of a retailer.

TD40

40- North 227 A- CC

Joint Session ENRE-Env & Sustainability/Energy:

Environmental Energy Models

Sponsor: Energy, Natural Res & the Envi/ Environment and

Sustainability & Energy, Natural Res & the Environment/Energy

Sponsored Session

Chair: Cheng-Marshal Wang, Environment Canada, 10 Wellington

Street, 25th Floor, Gatineau, QC, K1A0H3, Canada,

[email protected]

1 - Carbon Mitigation and Competition

Zachary Gillerlain, University of Cincinnati,

2925 Campus Green Dr, Cincinnati, OH, 45221,

United States of America, [email protected]

We consider two firms in a single product environment. Using stylized models, we seek to determine the optimal number of stores for two competing firms in a carbon-regulated environment. We quantify the carbon impact of competition and we compare the actions of the firms to the actions that would be employed by a social planner under several carbon mitigating policies.

2 - A Stochastic Mixed-integer Programming Approach to Facility and Energy-technology Management

Stephen Stoyan, Analytics Manager, Starbucks Coffee Company,

2401 Utah Street S., Seattle, WA, 98134, United States of America, [email protected], Maged Dessouky

We present a model that addresses facility location and employs energy technologies that aim to improve a companies carbon footprint. The approach involves a Stochastic Mixed-Integer Program (SMIP) that minimizes cost, emission levels, etc. while also considering the optimal location in the companies network. The results provide encouraging outcomes with respect to cost, emission levels, and energy-technologies that should be utilized for future generation.

3 - A Combined Model of Classical and Impulse Controls for

Emission and Stock Abatement Policies

Akira Maeda, The University of Tokyo, [email protected], Motoh Tsujimura

This study investigates emission and stock abatement policy decisions by the use of a stochastic optimal control model. We begin with our discussion by applying an absolute continuous control to emission flow abatement and management.

Then, we extend the model to a direct control of carbon stock. We introduce a specific feature of the cost structure for stock control and assume the cost function is described as a discontinuous and subadditive one. For this extension, we use an impulse control.

4 - The New Energy Model Design on Oil Sand Production and Export

Cheng-Marshal Wang, Environment Canada,

10 Wellington Street, 25th Floor, Gatineau, QC, K1A0H3, Canada,

[email protected], Nick Macaluso

International energy trading is of vital importance to Canada’s continued economic growth. The uncertainty on Keystone XL and Gateway pipelines will affect the projection of Canadian oil sand export and production. We will develop a two-stage Stochastic Programming EC-IAM (Integrated Assessment Model) to explore the implications of energy productions and export. The Hoteling framework on modeling depletable resources will be reviewed and refined to better project Oil Sand production.

352

INFORMS Phoenix – 2012

TD41

41- North 227 B- CC

Stochastic Programming Models for Wildfire

Response Planning

Sponsor: Energy, Natural Res & the Environment/Forestry

Sponsored Session

Chair: Lewis Ntaimo, Associate Professor, Texas A&M University,

3131, College Station, TX, 77843, United States of America, [email protected]

Co-Chair: Julian Gallego, Texas A&M University, Enemerging

Technologies Building, Industrial & Systems Engineering, College

Station, TX, TX 77843-3, United States of America, [email protected]

1 - A Simulation-optimization Approach for Large-scale Wildfire

Extended Attack Response Planning

Michelle M. Alvarado, PhD Student, Texas A&M University,

3131 TAMU, College Station, TX, 77843, United States of America, [email protected], Lewis Ntaimo

Our simulation-optimization approach integrates a stochastic integer programming (SIP) model for wildfire extended attack planning with DEVS-FIRE, a fire spread and suppression discrete event simulation model. The SIP model optimizes the location and timing of the deployment of firefighting resources to an escaped large-scale wildfire. The SIP decisions and weather forecasts are used in DEVS-FIRE to generate fire spread scenarios, which become input for the SIP model in the next decision period.

2 - A Probabilistically Constrained Programming Model for Wildfire

Initial Attack Planning

Julian Gallego, Texas A&M University, Enemerging Technologies

Building, Industrial & Systems Engineering, College Station, TX,

TX 77843-3, United States of America, [email protected],

Lewis Ntaimo

We propose a stochastic integer programming standard response model for initial attack planning as an alternative to other methods in literature. Risk is incorporated in the model by imposing probabilistic constraints on the minimum standard production rate required at a fire. We report results based on one of the

Texas Forest Service fire planning units in Texas.

3 - A Stochastic Integer Programming Model for Optimizing

Suppression Resource Assignments on a Wildfire

Erin McCowen, Colorado State University, Forest and Rangeland

Stewardship, 1472 Campus Delivery, Ft. Collins, CO, 80523-1472,

United States of America, [email protected], Mike Bevers,

Yu Wei

Assigning wildfire suppression resources safely and efficiently is a complex task.

We developed a spatial stochastic programming formulation to model fire spread, intensity and suppression activities. Solutions to the model suggest optimal suppression activities for a known ignition given a probability distribution of random weather scenarios. The formulation is designed to handle multi-objective fire policies where portions of a fire may be allowed to spread and other portions are contained.

4 - Fire Season Suppression Resource Planning using

Stochastic Programming

Yu Wei, Associate Professor, Colorado State University, Forest and

Rangeland Stewardship, 1472 Campus Delivery, Ft. Collins, CO,

80523-1472, United States of America, [email protected],

Mike Bevers

We simulated growth of more than 1,000 historical wildfires for a fire planning unit using hourly fire weather data. Simulation results provided spatially realistic fire perimeter and area parameters for scenarios in an otherwise aspatial multistage stochastic integer programming fire containment model. Model solutions suggest optimal fire suppression resource stationing and dispatch decisions for a fire season. Results of an initial test case are reported for the Black

Hills of South Dakota.

TD42

TD42

42- North 227 C- CC

Transportation Operations

Contributed Session

Chair: Seyed Behzad Aghdashi, PhD Candidate, North Carolina State

University, 2510-203 Avent Ferry Rd, Raleigh, United States of America

1 - The Downhill Plow Problem with Multiple Plows

Benjamin Dussault, University of Maryland, Department of

Mathematics, College Park, MD, United States of America, [email protected], Bruce Golden, Edward Wasil

A common problem in winter is determining snow plow routes. We allow for multiple plows and observe that, on steep streets, it is difficult to plow uphill. This multiple plow problem seeks to minimize the maximum route length. We present a heuristic that generates solutions close to a lower bound.

2 - Capacity Optimization of Isolated Intersections Embedded with Sorting Areas

Chiwei Yan, Tsinghua University, Department of Industrial

Engineering, Beijing, China, [email protected], Hai Jiang,

Siyang Xie

In this research, we study a new urban traffic control strategy named “sorting area” in order to increase capacity of intersections with over-saturated arms. A

Binary-Mixed-Integer-Linear-Programming (BMILP) model is established to get optimial signal settings and lane configurations on intersections embedded with sorting area. Numerical experiments are conducted to draw managerial insights.

A simulation model is also developed to validate the model results.

3 - Inefficiency of the User Equilibrium in Congestion Games under

Stochastic Demand

Chenlan Wang, Warwick Business School & DIMAP, Warwick

Business School, The University of Warwick, Coventry, CV4 7AL,

United Kingdom, [email protected], Bo Chen,

Xuan Vinh Doan

We study the degradation in network performance caused by travellers’ selfish behavior in a stochastic network environment. We present user equilibrium and system optimum conditions in non-atomic congestion games under stochastic demand on the basis of route choice model. Price of anarchy is adopted as the quantitative measure to study the inefficiency of the user equilibria under stochastic demand.

4 - Calibrating Traffic Flow Models through a Particle Filter

Resemble Model using Time-series Detector

Yang Lu, University of Maryland, College Park, MD, 20740,

United States of America, [email protected], Ali Haghani

The calibration of traffic flow models is a challenging task because analytical relationship between parameters and objective value is usually unobtainable. The computational complexity of GA method grows linearly with both the number of parameters and amount of field data used. This study proposed a particle filter resemble model to improve the computational efficiency of the parameter calibration process.

5 - Maximizing Traffic Flow Across a Freeway Facility using

Ramp Metering

Behzad Aghdashi, Research Assistant, North Carolina State

University, 2510-203 Avent Ferry Rd, Raleigh, NC,

United States of America, saghda[email protected]

In this research a flexible mathematical model has been developed for modeling traffic across a freeway facility. This flexibility makes models capable of modeling and optimizing different demand management strategies such as ramp metering, managed lanes, and hard shoulder running. Stochastic nature of the freeway capacity has been incorporated in the proposed mathematical models. Finally a linear programming model is proposed for this purpose.

353

TD43

INFORMS Phoenix – 2012

TD43

43- North 228 A- CC

State-of-the-art Implementation of OR Applications

Sponsor: Railway Applications

Sponsored Session

Chair: Marc Meketon, Vice President, Oliver Wyman, 1 University Sq.,

Suite 100, Princeton, NJ, 08540, United States of America, [email protected]

1 - Implement OR Model-driven Information Systems at

Norfolk Southern

Clark Cheng, Director Operations Research, Norfolk Southern

Corporation, 1200 Peachtree St NE, MS 12-117, Atlanta, GA,

30309, United States of America, [email protected],

Jignesh Patel

Operations research (OR) models are generally complicated by nature and therefore they are hard to understand and difficult to use for business users. In this presentation, we’ll discuss how to combine operations research and information technology to develop strategic and tactical information systems. The goal is to provide practical and usable tools to assist management with better decision-making, enhance business competitive advantage, and improve operating efficiency.

2 - Unsung Heroes for MultiRail Development

Marc Meketon, Vice President, Oliver Wyman, 1 University Sq.,

Suite 100, Princeton, NJ, 08540, United States of America, [email protected]

Systems for optimization of railway operating plans rely on a combination of algorithms, well-thought user interfaces, advanced software concepts, and hardware architectural designs. In this talk we discuss the contributions of these unsung heroes on advanced architecture, database design, parallelization, and user interfaces on the development of MultiRail.

3 - Interactive Decision Support Systems for Better Railroad

Planning

Ravindra Ahuja, President & CEO, Innovative Scheduling, GTEC,

2153 Hawthorne Road, Suite 128, Gainesville, FL, 32641,

United States of America, [email protected]

Service design teams at railroads create and maintain operating plans (blocking and train plans). But the actual execution may differ from the plan. In this presentation, we will give an overview and demonstration of a web-based, mapbased interactive decision support system to analyze the actual execution from shipment, train, block, and terminal perspectives, compare with the plan, and perform root cause analysis of any variance.

4 - BNSF Visualization Studio

Rachel Salvagio, Manager Network Analysis, BNSF Railway,

2600 Lou Menk Drive, Ft. Worth, TX, United States of America,

[email protected], Ron Griffith

With velocity being a constant focus at BNSF Railway, we recently added a component within our Enhanced Service Design (ESD) tool called Visualization

Studio that provides design managers information on velocity from a division level and waybill/car level. The product’s powerful visual depiction allows users to quickly identify opportunities for improvement.

TD44

44- North 228 B- CC

Supply Chain, Practice and Empirics

Contributed Session

Chair: Jesus Jimenez, Assistant Professor, Texas State University-

San Marcos, 601 University Dr, San Marcos, TX, 78666,

United States of America, [email protected]

1 - The Growing Price

Jing-An Li, Academy of Mathematics and Systems Science,

Chinese Academy of Sciences, No. 55, Zhongguancun East Road,

Haidian, Beijing, 100190, China, [email protected]

Recently, more and more cases are reported and questioned that the retail price of some vegetables is several times of the wholesale price. Produced by the manufacturer, wholesaled by the vendor, sold by the retailer, the vegetable’s price is gradually growing. Using the knowledge of supply chain management, this paper will analyze this scenario, and will discover the secret of the growth.

2 - Supply Network Architectures: Antecedents and Consequences

Myung Kyo Kim, Doctoral Candidate, Michigan State University,

North Business College Complex, 632 Bogue St. Rm N461,

East Lansing, MI, 48824, United States of America, [email protected], Ram Narasimhan

A supply chain is an excellent example of a multi-level complex system which has a strict architecture in that it includes coordination and collaboration with multiple partners, thus supply chain managers behave as “network architects” by defining the objectives and designating member companies of the network.

Responding to the needs for adopting network perspective to SCM, this study empirically investigates the antecedents and performance consequences of different supply network architectures.

3 - Has the Bullwhip Effect been Underestimated? - A Study

Linking Customers and Suppliers

Olov Isaksson, Doctoral Candidate, EPFL, Station 5, ODY, 4.15,

Lausanne, 1025, Switzerland, [email protected], Ralf Seifert,

Pinar Uysal

We collect firm-level data that links customers with suppliers and estimate the size of the bullwhip effect. Contrary to previous studies, we find a significant bullwhip effect across industries. Our findings also show the bullwhip effect to be significantly higher than previously estimated. We investigate potential reasons for these differences.

4 - Variables Affecting the Semiconductor Manufacturing Supply

Chain in the 450mm Wafer Size Transition

Jesus Jimenez, Assistant Professor, Texas State University-San

Marcos, 601 University Dr, San Marcos, TX, 78666, United States of America, [email protected], Garret Wilson

The semiconductor manufacturing industry is transitioning to 450mm wafers.

There are several strategic-decision variables of the wafer size transition that will impact the entire semiconductor manufacturing supply chain; these variables include cycle time improvement, wafer transition cost, paying the bill, return on investment, productivity, and equipment/material. Empirical and simulation models were used in this research to evaluate the effects of these variables and provide recommendations.

TD45

45- North 229 A- CC

Joint Session JFIG/INFORM-ED; Maximizing Learning

Impact with Web-Based Simulation and Classroom

Activities

Sponsor: Junior Faculty Interest Group & INFORM-ED

Sponsored Session

Chair: Elizabeth Durango-Cohen, Illinois Institute of Technology,

565 W. Adams, Chicago, IL, United States of America, [email protected]

1 - Using Simulation Software to Drive Conceptual Understanding in Undergraduate OM Course

Elizabeth Durango-Cohen, Illinois Institute of Technology,

565 W. Adams, Chicago, IL, United States of America, [email protected]

In this talk, we’ll discuss various simulation packages, including Littlefield

Technologies, the EBeer game as well as the use of clickers, and their impact on student understanding and participation.

2 - Classroom Activities to Demonstrate Various Topics in OM/SCM

Ozgun Caliskan Demirag, Pennsylvania State University- Erie,

The Behrend College, 5101 Jordan Road, Erie, PA, 16563,

United States of America, [email protected], Berrin Aytac

This talk will present a number of classroom activities that can be used in teaching various OM/SCM topics. The activities are designed as relatively simple, shortduration, easy-to-prepare, and fun class exercises, which can be used to explain introductory concepts as well as clarify some of the more complex topics in the relevant areas.

3 - Online Games in Undergraduate Introductory OM Courses

Sam Wood, Responsive Learning Technologies, 4546 El Camino

Real, #243, Los Altos, CA, 95014, United States of America, [email protected]

Littlefield is a competitive online simulation used to teach introductory OM topics like process analysis and inventory control. Unlike the case for electives and graduate courses, big undergraduate sections of required OM courses appear to present special challenges when using online games like Littlefield. Using faculty interviews and data from actual Littlefield games over a four-year period we identify some specific challenges and test ways of addressing them.

354

TD46

46- North 229 B- CC

Game Theory: Application

Contributed Session

Chair: Monireh Mahmoudi, PhD Student, Wayne State University,

4815 Fourth Street, Department of Industrial & Systems Eng, Detroit,

MI, 48202, United States of America, [email protected]

1 - A Game between a Government and a Tobacco Manufacturer

Md Ahmed, Research Assistant, University at Buffalo, SUNY,

Buffalo, NY, 14226, United States of America, [email protected], Changhyun Kwon, Jun Zhuang

Rice is the staple food of nearly half of the world’s population. Ensuring domestic supply of rice from outside sources is difficult. In order to ensure food security, governments provide subsidy in agriculture. Sometimes, public money used for subsidy goes toward promoting undesirable crops like tobacco. We considers a game between a govt. and a tobacco manufacturer in which the govt. decides on a mix of subsidies and the manufacturer decides on declaring a purchase price of tobacco.

2 - A Mixed Leadership Game of Cooperative Advertising between a Manufacturer and a Retailer

Anshuman Chutani, Assistant Professor of Supply Chain

Management (Visiting), School of Management, Binghamton

University, P.O. Box 6000, Binghamton, NY, 13902, United States of America, [email protected], Alain Bensoussan,

Suresh Sethi, Shaokuan Chen

We study a class of sequential differential games where decision makers act both as leaders and followers depending on the nature of the decision. We develop feedback solutions in these mixed leadership differential games. We study an example of cooperative advertising between a manufacturer and a retailer, who decide national and local advertising rates, respectively. Each party first declares its share of other’s advertising expenditure, and then both decide their respective advertising rates.

3 - A Game Theoretic Approach to Analyze Rear-ending Crashes

Indrajit Chatterjee, University of Minnesota, 500 Pillsbury Drive

S.E, Minneapolis, MN, 55455-0116, United States of America, [email protected]

Analyzing rear-ending crashes requires understanding driver behaviors in a typical braking-to-stop scenario. Such interactions between leading and following vehicle can be viewed as a product of evolutionary game, where a population of drivers is randomly paired in a two-player game. A stable long-run distribution of driving strategies is the outcome of the evolutionary process, which in-turn can be used to estimate the frequency of rear-ending crashes on congested freeways.

4 - Towards a Counterfeit-proof Global Supply Chain

Morteza Pourakbar, Rotterdam School of Management,

Erasmus University, Burg Oudlaan 50, Rotterdam, Netherlands, [email protected], Rob Zuidwijk

Counterfeiting activities are known as a major harm to legitimate supply chains.

In this study, we investigate the role of customs authorities and rights-owners in the fight against counterfeiting. In a game setting, we analyze how customs and rights-owner alliance results in lower levels of counterfeiting risks.

5 - Competitive Pricing Dynamics under a Price Leader:

Dairy Industry Case

Monireh Mahmoudi, PhD Student, Wayne State University, 4815

Fourth Street, Department of Industrial & Systems Eng, Detroit,

MI, 48202, United States of America, [email protected],

Ratna Babu Chinnam

Data from five major dairy product firms were used to determine pricing status of ultrafiltration cheese. The leader firm claims to always sell the product near the ceiling price. We test the claims of the leader by calculating the optimal prices for all the firms in four different states using game theory principles. The price in each period was also predicted using a time series approach and compared to results from the game theory approach. The analysis supports the claim of the leader firm.

INFORMS Phoenix – 2012

TD48

TD47

47- North 230- CC

Route Choice and Traffic Assignment

Sponsor: Transportation Science & Logistics/ Intelligent

Transportation Systems (ITS)

Sponsored Session

Chair: Jing Ding, Research Assistant, University of Massachusetts

Amherst, 130 Natural Resources Rd, Amherst, MA, 01003,

United States of America, [email protected]

1 - Adaptive Transit Routing under Uncertainty

Stephen Boyles, Assistant Professor, The University of Texas,

Austin, 1 University Station C1761, Austin, TX, United States of

America, [email protected], Tarun Rambha, Travis Waller

An adaptive transit routing (ATR) problem on a stochastic transit network is defined and is formulated as a Markov Decision Process. State space reduction through pre-processing is achieved by solving variants of time dependent shortest paths and applying rules of dominance. Sampling based approximation techniques are used to solve an instance of the problem to demonstrate its application.

2 - Vacant Taxi Driver Routing Behavior Learning Process

Xianbiao Hu, Research Assistant, University of Arizona,

1209 E. 2nd Street, Tucson, AZ, 85721, United States of America, [email protected], Yi-Chang Chiu

We propose a vacant taxi driver’s routing behaviour model, within which a taxi driver determines his/her optimal route based on the experience traffic condition and customer arrival rate, but also learn from the actual realization. By the end of day taxi driver gains actual experience about the traffic situation and customer distribution, compares it to the previous experience, then forms updated knowledge for the next day. Numerical convergence analysis demonstrates the property of such a learning process.

3 - Hyperpath Equilibrium Models to Simulate Adaptive

Routing Behavior

Stephen Boyles, Assistant Professor, The University of Texas,

Austin, 1 University Station C1761, Austin, TX, United States of

America, [email protected], Ravi Venkatraman, Avinash

Unnikrishnan, Rachel James

An interactive online game is conducted to study adaptive routing behavior under the conditions of complete en-route information. In particular, this game aims to see how well drivers can use information, compared to the theoretical optimum value based on an online shortest path algorithm. Several behavioral contexts are studied.

4 - Adaptive Route Choice Models: Specification, Choice Set

Generation, and Estimation

Jing Ding, Research Assistant, University of Massachusetts

Amherst, 130 Natural Resources Rd, Amherst, MA, 01003,

United States of America, [email protected]

We develop a modelling framework for adaptive route choice based on vehicle trajectory data in a general stochastic time-dependent network, here a traveler could revise route choice based upon real-time traffic information. We demonstrate the model specification, choice set generation algorithm, and estimation process in a large real-life network using synthetic data.

TD48

48- North 231 A- CC

Facility Logistics III

Sponsor: Transportation Science & Logistics/ Facility Logistics

Sponsored Session

Chair: Ananth Krishnamurthy, University of Wisconsin-Madison, 1513

University Avenue, ME 3258, Madison, WI, United States of America, [email protected]

1 - Solving the Reshuffling Problem in a Distribution Center

Jennifer Pazour, Assistant Professor, University of Central Florida,

4000 Central Florida Blvd, Orlando, FL, 32816, United States of

America, [email protected], Hector Carlo

Over time, product demand profiles change; consequently, the assignment of products to locations in a distribution center also changes over time. We study the reshuffling problem, which occurs when products need to be rearranged in a distribution center. We develop a mathematical model and solution approach to determine the movement sequence that should be taken to transform an initial layout to a desired, final layout that minimizing travel time costs.

355

TD49

2 - Integrated Design of Order Picking and Sortation Systems

Fahrettin Eldemir, Assistant Professor, Yildiz Technical University,

Industrial Engineering Dept, Yildiz, Besiktas, Istanbul, Turkey, [email protected], Recep Kizilaslan, Elif Karakaya

Order picking and order sorting are the two main operations that are mostly encountered within a warehouse. The research in the literature generally addresses the operational issues considering the both system as independent systems. In this study a performance evaluation model is suggested for the integration of the Order Picking Systems and Order Sortation Systems. Detailed aspects of design and operational parameters are addressed.

3 - Comprehensive Tactical Supply Chain Planning under Uncertainty

Aly Megahed, PhD Student, H. Milton Stewart School of Systems and Industrial Engineering, Georgia Institute of Technology, 765

Ferst Drive NW, Atlanta, GA, 30332, United States of America, [email protected], Marc Goetschalckx

We present a comprehensive two-stage stochastic supply chain tactical planning model for multi-period, multi-commodity, and multi-echelon supply chains. The first stage decisions include supplier selection and supplier contract design. The second stage decisions include inventory, backorder, transportation, product flow quantities and resource utilizations. The solution methodology and numerical results will be shared.

4 - Layout Design for Relief Centers in Disaster Affected Areas

Ananth Krishnamurthy, University of Wisconsin-Madison,

1513 University Avenue, ME 3258, Madison, WI,

United States of America, [email protected]

We use queuing models to analyze alternate layouts for relief centers that distribute relief aid to affected victims at a disaster affected region. The alternate layouts in the study integrate knowledge from the traffic control literature and we investigate the joint impact of the service needs of victims and accessibility constraints at the site.

TD49

49- North 231 B- CC

INFORMS Phoenix – 2012

Pricing and Financing of Transportation Systems: I

Sponsor: Transportation Science & Logistics

Sponsored Session

Chair: Yafeng Yin, Associate Professor, University of Florida,

365 Weil Hall, Gainesville, FL, 32611, United States of America, [email protected]

1 - Managing Morning Commute with Tradable Mobility Credits

Marco Nie, Northwestern University, Evanston, IL, 60208,

United States of America, [email protected]

We assume that the congestion management authority will delineate a peak-time window and require all traveler who pass the bottleneck within that window to either pay certain units of mobility credits, or a peak time toll (which is comparably more expensive). The authority rewards credits to anyone who passes the bottleneck during the designated off-peak time window. An artificial market will be created so that the users may trade these credits with each other.

2 - Pricing Policy in a Stackelberg Game on Supply Chain Network with Imperfect Information

Tao Yao, Pennsylvania State University, 349 Leonhard Building,

University Park, PA, 16802, United States of America, [email protected], Amir Meimand, Terry Friesz

In this paper we present a model for stackelberg game on supply chain network in which supplier is a leader and retailers are followers and have uncertain observability of leader’s policy which is the whole sale price. Then we study the effect of this uncertainty on the profitability of supplier and will discuss how different pricing policies can mitigate the negative effect of uncertainty.

3 - Optimal Dynamic Pricing for Parking Management

Sean Qian, Stanford University, Department of Civil &

Environmental Eng., Stanford, CA, 94305, United States of

America, [email protected], Ram Rajagopal

The objective of this research is to maximize the benifits of parking management by optimal pricing and sensing. A generic parking model is presented for a set of sequential parking lots. We propose the system optimum and the minimum congestion parking pricing schemes. The model can be implemented in practice by utilizing parking sensors to set optimal on-line parking prices.

4 - Equity and Efficiency Analysis of Vehicle Usage Restriction,

Ownership Quota, and Pricing Policies

Shanjiang Zhu, Assistant Research Scientist, University of

Maryland, 1173 Glenn L. Martin Hall, College Park, MD, 20742,

United States of America, [email protected], Longyuan Du,

Lei Zhang

This research develops an analytical framework for analyzing and comparing transportation rationing policies, which consists of a mathematical model of joint household vehicle ownership and usage decisions and welfare analysis methods based on compensating variation and consumer surplus. This research extends previous research to consider multiple user groups on a network.

TD50

50- North 231 C- CC

Telecommunications

Contributed Session

Chair: Liang Zheng, Researcher, IBM Research - China, Diamond

Building 19-A, Zhongguancun Software Park, Beijing, 100193, China, [email protected]

1 - Optimal Relay Node Placement in Sensor Networks using Node

Cut Inequalities- A Branch & Cut Appraoch

Ashutosh Nigam, Doctoral Student, IIM Lucknow, FPM 28, IIM

Lucknow, Lucknow, 226013, India, [email protected]

Objective is to find a minimal set of relay nodes which facilitate communication between all source nodes (which sense the signals) and a root node. We provide a path based formulation and then discuss the concept of node cuts (set of relay nodes, deletion of which disconnects the source node and the root node) using which we propose an alternate projection formulation to solve the problem more efficiently. Facet defining conditions for the node cuts are also proved in the work.

2 - A Comparative Analysis of Steiner Ring Star

Problem Formulations

Junsang Yuh, Korea University, Anamdong 5 Ga, Sungbuk Gu,

Seoul, Korea, Republic of, [email protected], Youngho Lee,

Gigyoung Park

We deal with the Steiner ring star problem arising from the design of telecommunication networks. We develop several mixed 0-1 integer programming models for subtour elimination constraints. By implementing the reformulation-linearization technique (RLT), we devise valid inequalities tightening the LP relaxation. Computational results compare the LP relaxation lower bounds and shows the effectiveness of the RLT procedures.

3 - Optimal Relay Selection Algorithm for Wireless Sensor

Networks

Liang Zheng, Researcher, IBM Research-China, Diamond Building

19-A, Zhongguancun Software Park, Beijing, 100193, China, [email protected], Hai Rong Lv, Jun Hua Ma,

Wen Jun Yin, Jin Dong, Wei Zhao, Jin Feng

The paper issues an innovative algorithm to select relays in cooperative communication. The relay selection scheme can not only get quality QoS but also reduces the complexity of system, which is based on the eigenvalue decomposition. The simulation shows that compared wth direct communication, our method significantly improve the channel capacity and save the transmit power.

4 - Optimizing Fiber Delay Line Banks in Multi-channel

Optical Switches

Yavuz Gunalay, Bahcesehir University, Faculty of Economics and Admin. Science, Besiktas, Istanbul 34353, Turkey, [email protected], Naik Akar

An optical packet switching system which has N input-output fibers and W wavelength channels for each fiber is considered. In such systems, fiber delay lines (FDL) can be deployed to resolve contention. The size of FDL bank is optimized for a given level of QoS.

356

TD51

51- North 232 A- CC

Joint Session: MAS/SPPSN/CPMS/SS: Tutorial:

Transforming Military Supply Chains - Management

Innovation in DoD

Sponsor: Military Applications, CPMS, The Practice Section, &

Public Programs, Service and Needs

Sponsored Session

Chair: Greg Parlier, Colonel, USA Retired, Institute for Defense

Analyses, 255 Avian Lane, Madison, AL, United States of America, [email protected]

1 - Tutorial: Transforming Military Supply Chains - Management

Innovation in DoD

Greg Parlier, Colonel, USA Retired, Institute for Defense Analyses,

255 Avian Lane, Madison, AL, United States of America, [email protected]

This tutorial offers a practical approach for understanding the Army’s extremely complex global logistics system. The new concept of “management innovation as a strategic technology” (MIST) is introduced and described. Cutting-edge supply chain theory and powerful analytical methods are applied to this seemingly intractable national security resource challenge which has remained on the

Government Accountability Office’s (GAO) “high-risk” list for two decades now.

TD52

INFORMS Phoenix – 2012

52- North 232 B- CC

Joint Session ENRE Energy/Environment &

Sustainability: Biomass Energy Market and

Supply Chain

Sponsor: Energy, Natural Res & the Environment/Energy & Energy,

Natural Res & the Envi/ Environment and Sustainability

Sponsored Session

Chair: Lizhi Wang, Iowa State University, 3016 Black Engineering,

Ames, IA, 50011, United States of America, [email protected]

1 - Renewable Energy and Competition for Biomass: Implications for Land use and Food Prices

Hayri Onal, University of Illinois at Urbana-Champaign,

305 Mumford Hall, Urbana, IL, 61801, United States of America, [email protected], Xiaoguang Chen

Recent US energy policies aim to substitute fossil fuels with renewable fuels. The

RFS relies largely on biofuels produced from biomass, RPS requires some of the power generation to come from biomass in addition to solar, wind, hydro, and geothermal sources. We present a large scale nonlinear MIP model to simulate the competition between RFS & RPS for limited biomass and determine the spatial layout of biomass production and location of processing facilities. Empirical results are presented.

2 - Competitive Supply Chain Design for the Emerging

Biofuel Industry

Yun Bai, University of Illinois at Urbana-Champaign, 205 N.

Mathews Avenue, Urbana, IL, United States of America, [email protected], Jong-Shi Pang, Yanfeng Ouyang

We study a competitive supply chain design problem for an emerging industry which competes for resource supply with existing industries. A Stackelberg leader-follower game model is proposed to determine optimal facility location, scouring and pricing for the emerging industry, resource allocation for multiple suppliers. Lagrangian relaxation based solution approach is developed to solve the

DC-MPEC model efficiently for large scale problem instances.

3 - Optimization of Agricultural Biomass Supply Chain for

Cellulosic Ethanol Production

Taraneh Sowlati, University of British Columbia, Department of

Wood Science, Faculty of Forestry, Vancouver, BC, V6T 1Z4,

Canada, [email protected], Mahmood Ebadian,

Shahab Sokhansanj

In this paper, a combined simulation/optimization model is developed to model and analyse Satellite Storages (SSs) in an agricultural biomass supply chain for cellulosic ethanol production. The time dependence and stochastic nature of the supply chain are taken into account using the simulation model. The optimization model addresses the optimum number and location of SSs. Three storage scenarios are identified and their efficiency are compared.

4 - Potential Competition of Biomass for Bioelectricity and Biofuel under Renewable Portfolio Standards

Lizhi Wang, Iowa State University, 3016 Black Engineering,

Ames, IA, 50011, United States of America, [email protected],

Mohammad Rahdar, Guiping Hu

We study the potential competition of biomass for bioelectricity and biofuel under the renewable fuel standard and renewable portfolio standards. Supply functions for biomass and other renewable energy sources are used to assess the availability of renewable energy resources; demand of renewable energy is estimated from

EIA energy outlook, RFS2, RPS, and RPG.

TD53

53- North 232 C- CC

Pricing in Electricity Markets Including

Nonconvexities

Sponsor: Energy, Natural Res & the Environment/Energy

Sponsored Session

TD53

Chair: Antonio Conejo, Professor, Universidad Castilla - La Mancha,

Campus Universitario, Ciudad Real, 13071, Spain,

[email protected]

1 - Do Centrally Committed Markets Provide Useful Price Signals?

Ramteen Sioshansi, Assistant Professor, The Ohio State University,

240 Baker Systems, 1971 Neil Avenue, Columbus, OH, 43215,

United States of America, [email protected], Ashlin Tignor

The promise of restructured electricity markets is that prices, that are computed in a day-ahead unit commitment, provide signals for generation and transmission to enter and exit the market. We show that the prices generated by a centrally committed market are highly sensitive to which near-optimal solution is used, resulting in variable generator profits. We also show that peaking generators and transmission zones most prone to congestion are most likely to experience such variability.

2 - An Extreme-point Global Optimization Technique for Computing

Convex Hull Prices

Gui Wang, University of Illinois at Urbana-Champaign,

206 Transportation Building, Urbana, IL, United States of America, [email protected], Uday Shanbhag, Tongxin Zheng,

Eugene Litvinov, Sean Meyn

We present an extreme-point-based algorithm for obtaining convex-hull electricity prices that relies on globally maximizing the Lagrangian dual of an

MILP. Two practical limitations common to such methods are the requirement of complete subgradient information and zigzagging behavior close to the solution.

The former was handled by avoiding explicit enumeration while the latter was mitigated by introducing an alternative merit function.

3 - Pricing of Nonconvexity – A Forward Capacity

Market Experience

Tongxin Zheng, Technical Manager, ISO New England,

1 Sullivan Road, Holyoke, MA, 01106, United States of America, [email protected], Eugene Litvinov

Defining prices from a nonconvexity optimization problem is a very important issue in the electricity market with the lumpy resources. This paper presents a nonconvex forward capacity auction problem that minimizes the total consumer payment, as well as a zero uplift pricing scheme that is adopted in the forward capacity market in ISO New England under a zonal configuration.

4 - Finding Better Pricing Strategies for Electricity

Mingguo Hong, Midwest ISO, 5407 Alvamar Place, Carmel, IN,

46033, United States of America, [email protected]

The Location Marginal Price (LMP) based electricity market design has an inherent weakness in that it cannot guarantee the recovery of a generator’s startup and no load costs. As a result, the Regional Transmission Organization

(RTO) would have to provide make-whole payment in settlement. RTOs in North

America have made various design changes to improve pricing. This talk reviews the market design changes. It intends to reveal issues with LMP pricing and solicit alternative problem solutions.

357

TD54

INFORMS Phoenix – 2012

TD54

54- Regency Ballroom A- Hyatt

Applications in Open Pit Mining

Sponsor: Energy, Natural Res & the Environment/Mining

Sponsored Session

Chair: Alexandra Newman, Associate Professor,

Colorado School of Mines, 1500 Illinois Street, Golden, CO, 80401,

United States of America, [email protected]

1 - Long-term Planning for an Underground Mine

Michel Gamache, Professor, Polytechnique Montrèal, C.P. 6079 succ. Centre-Ville, Montreal, QC, H3C3A7, Canada, [email protected], Jean Collard

In this presentation, we will present a model of long-term planning for Raglan

Mine, a large nickel mining complex in the Nunavik region at the extreme limit of Northern Quebec, Canada. We will discuss the mixed integer linear programming model and the constraints that are related to the various operating characteristics of this mine in the sub-Arctic Circle. Results of solutions will be presented.

2 - A Study of the Bienstock-Zuckerberg Algorithm: Incorporating

Price Uncertainty

Marcos Goycoolea, School of Business, Universidad Adolfo Ibañez,

Santiago, Chile, [email protected], Daniel Espinoza,

Eduardo Moreno, Gonzalo Muñoz, Maurice Queyranne

A major concern of mine planners is the volatility of ore prices. Ben-tal and

Nemirovski propose a robust optimization model for dealing with this uncertainty that reduces to solving a second-order-cone programming problem (SOCP). We describe an extension of the Bienstock-Zuckergerg algorithm for the precedence constrained production scheduling problem (PCPSP) that incorporates uncertainty by using this robust optimization paradigm. Computational results on real data sets are presented.

3 - Evaluating Alternative Formulations for Open-pit Mine

Block Sequencing

Thomas Vossen, University of Colorado at Boulder, 995 Regent Dr,

Boulder, CO, United States of America, [email protected],

Jose Ramirez

The objective of open pit mine block sequencing is to determine a feasible extraction schedule from an open pit mine that maximizes profits over a given planning period. We propose and evaluate alternative formulations for this wellknown problem that are derived from a non-linear programming formulation, and discuss preprocessing strategies that can significantly reduce the size of the resulting problems. Experimental results show the potential of our approach.

TD55

55 – Regency Ballroom B - Hyatt

Transformation in Healthcare Delivery Through

Industry-academic Cooperative Research and

Implementation

Sponsor: Health Applications Society

Sponsored Session

Chair: Eva Lee, Professor and Director, Industrial and Systems

Engineering Georgia Institute of Technology, Georgia Institute of

Technology, Atlanta GA, United States of America, [email protected]

1 - Reducing Surgical Site Infections

Eva Lee ,Professor and Director, Industrial and Systems

Engineering Georgia Institute of Technology, Georgia Institute of

Technology, Atlanta GA, United States of America, [email protected], Ling Ling

This project is joint with Grady Memorial Hospital. Surgical-site infections (SSI) are a national problem, occurring in an estimated 2.8% of all procedures (for some, it is up to 30%). It causes complication, prolong length and stay and unnecessary burden on patients, hospitals and 3-party payers. We describe our recent successes in reducing SSI for coronary artery bypass graft through systems modeling and process optimization.

2 - Appointment Access & Engineering Control of Healthcare

Systems

James Benneyan, Northeastern University, 360 Huntington Ave.,

334 Snell Engineering Center, Boston MA 02120, United States of

America, [email protected]

In this work, we investigate value of engineering feeback control to dynamically adjust staff levels or work hours in order to achieve any acceptable appointment access limit.

3 - Supporting Multi-disciplinary Team (MDT) Collaboration

Through the EMR

Harriet Nembhard, Pennsylvania State University, Philadelphia PA,

United States of America, [email protected]

This talk aims to understand the role of non-clinical team members in a multidisciplinary team and their utilization of electronic medical records. The work is interview-based and we will discuss some of our findings.

4 - Evaluating the Impact of Discharge Phone Calls

Larry Gamm, Texas A & M University, Heatlth Science Center,

Bryan TX, United States of America, [email protected]

We focus on evaluating the impact of Discharge Phone Calls have had on reducing readmission. We will discuss how the characteristics of these calls have in effecting the execution and their usefulness to patients and hospitals. We will also discuss the impact on actual hospital readmission rates through actual hospital data analysis.

TD56

56- Curtis A- Hyatt

Joint Session TMS/ORG: Outsourcing of Knowledge-

Based Tasks: Challenges, Strategies, Performance, and Implications

Sponsor: Technology Management & Organization Science

Sponsored Session

Chair: Saikat Chaudhuri, Assistant Professor of Management, The

Wharton School, University of Pennsylvania, 2029 Steinberg Hall-

Dietrich Hall, 3620 Locust Walk, Philadelphia, PA, 19104-6370,

United States of America, [email protected]

1 - Capability Development Across Firm Boundaries: Comparing

Offshore Outsourcing of R&D vs. IT Services

Saikat Chaudhuri, Assistant Professor of Management, The

Wharton School, University of Pennsylvania, 2029 Steinberg Hall-

Dietrich Hall, 3620 Locust Walk, Philadelphia, PA, 19104-6370,

United States of America, [email protected]

Motivated by globally disaggregating firms, we compare the performance drivers in the offshore outsourcing of more routinized, codified IT services with less routinized and codifiable R&D work, to identify the conditions under which more central tasks can be located outside. Our analysis of a sample of such projects by a leading vendor suggests that capability creation across firm boundaries is fruitful under certain organizational designs, bearing implications for notions of core vs.

periphery.

2 - Coordination, Contracts or Control? The Drivers of Quality In

Offshore Service Produciton

Ravi Aron, Assistant Professor, The Johns Hopkins Carey Business

School, 100 International Drive, Room 1331, Baltimore, MD,

21202, United States of America, [email protected],

Praveen Pathak, Ying Liu

We study the factors that drive the quality of output in services produced offshore. We investigate how the nature of work, the extent of control exerted by buyers and suppliers of offshore services and the features of the contract all impact on the quality of work produced offshore. Our research is based on a panel data set (a balanced panel of suppliers and buyers) from multiple countries and multiple buyers and suppliers of offshore services.

3 - The Architecture of Multi-partner Alliances in R&D Projects:

Scale, Ambidexterity and Integration

Anant Mishra, Assistant Professor, George Mason University,

4400 University Drive, Fairfax, VA, United States of America, [email protected], Alan MacCormack, Aravind Chandrasekaran

Why are some partnering alliances more successful than others? And how should firms structure their partnering alliances, particularly when such alliances involve multiple partners? Using primary data on multi-partner alliances across 147 R&D projects in six different industries, we develop and test hypotheses that examines the interrelationship between the different elements of partnering architecture and quality performance in an R&D project.

358

INFORMS Phoenix – 2012

TD57

57- Curtis B- Hyatt

Asymptotic Analysis of Large-Scale Service Systems

Sponsor: Applied Probability

Sponsored Session

Chair: Guodong Pang, Assistant Professor, Pennsylvania State

University, University Park, PA, United States of America, [email protected]

1 - Control of Patient Flow in Emergency Departments: Multiclass

Queues with Feedback and Deadlines

Junfei Huang, National University of Singapore, Singapore,

117592, Singapore, [email protected], Avishai Mandelbaum,

Boaz Carmeli

We consider the control of patient flow through physicians in emergency departments: a choice must be made between triage patients who are yet to be checked vs. those who are in-process (IP). Physicians’ capacity is modeled as a multi-class queueing system with feedback, in which triage patients face deadline constraints while IP patients incur delay costs. We propose and analyze a scheduling policy that minimizes total queueing cost while satisfying all constraints, asymptotically.

2 - On the Optimal Control of Matching Queues

Itai Gurvich, Northwestern University-Kellogg School of

Management, Evanston, IL, United States of America, [email protected], Amy Ward

We study a dynamic matching problem in which arriving jobs can leave the system only after being matched with one or multiple jobs of different types. In high-volume systems, a “matching pooling” condition, reminiscent of the

“resource pooling condition” in parallel server queues, facilitates a reduction of the original problem to an equivalent shortage formulation. We prove that a greedy policy based on repeatedly solving static matching problems is asymptotically optimal.

3 - A Logarithmic Safety Staffing Rule for Contact Centers with

Call Blending

Ohad Perry, Northwestern University, 2145 Sheridan Rd.,

Evanston, IL, United States of America, [email protected], Guodong Pang

We study a large-scale contact center that handles inbound and outbound calls.

Inbound calls arrive as an exogenous stream of customers with impatience, while outbound calls are modeled as an infinite queue of jobs, waiting to be processed.

The goal is to serve inbound calls subject to service level constraints while keeping the outbound throughput rate at a desired level. We show that a logarithmic safety staffing is sufficient to meet these objectives.

4 - Fluid Limits for Multiclass Many-server Queues with Reneging under Global FCFS Discipline

Weining Kang, University of Maryland at Baltimore County,

Baltimore, MD, United States of America, [email protected],

Guodong Pang

We consider a multiclass many-server queueing system, where customers from different classes are served according to a global first-come-first-serve discipline and will abandon from their queues when their patience times are reached. We establish the fluid limit for such a system under some mild assumptions on the general service and patience time distributions.

TD58

58- Phoenix East- Hyatt

Stochastic Models in Services and Computing

Sponsor: Applied Probability

Sponsored Session

Chair: Alan Scheller-Wolf, Carnegie Mellon University, 5000 Forbes

Avenue, Pittsburgh, PA, 15213, United States of America, [email protected]

1 - Exact Analysis of the M/M/2 with Setup Times and other

Hard Variants

Anshul Gandhi, Carnegie Mellon University, 5000 Forbes Avenue,

Pittsburgh, PA, 15213, United States of America, [email protected], Sherwin Doroudi, Alan Scheller-Wolf,

Mor Harchol-Balter

Many classic queueing problems have a simple Markov chain representation, yet solving this Markov chain is very difficult. An example is the M/M/2/Setup system where there is a setup time to turn a server on. In this work, we introduce a novel technique to analyze Markov chains that works by combining ideas from renewal reward theory and busy period analysis. We demonstrate this technique by solving the M/M/2/Setup and other similarly difficult variants.

TD59

2 - The Benefit of Introducing Variability in Quality Based

Service Domains

Ying Xu, Carnegie Mellon University, 5000 Forbes Avenue,

Pittsburgh, PA, 15213, United States of America, [email protected], Alan Scheller-Wolf, Katia Sycara

We consider a single-server queueing system in which the value customers obtain from service increases with service time but decreases with waiting time. We show that the system value can be improved by differentiating a homogeneous customer group into different service rates, even if the service rates are assigned independent of system state. Though such a differentiation service increases service time variance, the total waiting time can be reduced without affecting the total service time.

3 - Analysis of Border Crossing Stations

Katsunobu Sasanuma, PhD Student, Carnegie Mellon University,

Heinz College, Pittsburgh, PA, 15213, United States of America, [email protected], Robert Hampshire, Alan Scheller-Wolf

Border-crossing stations are often modeled using two connected birth-death

Markov chains. However, the performance analysis of a model of border-crossing stations is complicated. We propose a new method to analyze such connected

Markov chains. The method enables us to decompose Markov chains into basic sub-chains and makes our analysis simpler.

4 - Multi-priority M/M/c Queue

Jianfu Wang, PhD Candidate, University of Toronto, Rotman

School of Management, 105 St. George Street, Toronto, ON,

M5S3E6, Canada, [email protected],

Opher Baron

The multi-priority M/M/c queue has been widely studied. However its exact analysis was restricted to cases with identical service rates for different customer priorities. We use novel queueing analysis to provide closed form solution when service rates differ.

TD59

59- Phoenix West- Hyatt

Quantitative Risk Management

Sponsor: Applied Probability

Sponsored Session

Chair: Jeffrey Collamore, Associate Professor, University of

Copenhagen, Department of Mathematical Sciences, Universitetsparken

5, Copenhagen, DK-2100, Denmark, [email protected]

1 - Large Deviations for the Empirical Measure of an Importance

Sampling Algorithm

Henrik Hult, Associate Professor, Royal Institute of Technology,

Department of Mathematics, Stockholm, SE-100 44, Sweden, [email protected], Pierre Nyquist

Extreme risk in a financial or actuarial context is often quantified by a quantile or other tail-based risk measure. However, there may be a need to compute many rare-event probabilities for an accurate evaluation of a risk measure. In this talk we present a large deviations result for the empirical measure resulting from an importance sampling algorithm and show how the associated rate function can be used to quantify the performance of the algorithm.

2 - Nested Rare Event Simulation in Risk Management

Jie Xu, Assistant Professor, George Mason University, 4400

University Dr, MS 4A6, GMU Engr Bldg RM2100, Fairfax, VA,

22030, United States of America, [email protected],

Anand Vidyashankar

Nested simulation is an important tool in assessing actuarial and financial risks.

When the loss function involves extreme percentiles, use of Monte Carlo simulations within nested simulations will render the simulation solution inefficient. We address this problem by introducing a nested rare event simulation solution and describe the theoretical properties of our algorithm. We establish consistency and efficiency of our algorithm and provide a precise description of the simulation cost.

3 - Dynamic Network Models for Systematic Risk in Financial

Systems and Related Asymptotics

Anand Vidyashankar, Associate Professor, George Mason

University, Department of Statistics, 4400 University Drive, MS

4A7, Fairfax, VA, 22030, United States of America, [email protected]

Network models are being increasingly used to model and evaluate systematic risk in financial systems. However, it is known that systematic risk takes multiple forms and is dynamic in nature. To address this issue, we describe an evolving network model for financial systems and use network wide metrics(NWM) to evaluate the systematic risk. We provide a precise mathematical description of the asymptotic behavior of the NWM and their interpretations for the risk problem.

359

TD60

TD60

60- Remington- Hyatt

Airport Runway and Taxiway Scheduling

Sponsor: Aviation Applications

Sponsored Session

Chair: Karen Marais, Assistant Professor, Purdue University,

401 W. Stadium Avenue, ARMS 3225, West Lafayette, 47906,

United States of America, [email protected]

1 - End Around Taxiway Operational Optimization Approaches

Payuna Uday, Graduate Student, Purdue University,

401 W. Stadium Avenue, ARMS 3173, West Lafayette, IN, 47906,

United States of America, [email protected], Karen Marais

The adoption of end-around taxiways at airports with parallel runways has the potential to increase runway throughput, and reduce fuel burn and surface emissions. This study presents decision support models based on the environmental consequences of these taxiways at one airport. Two types of decision rules are evaluated (a) static rules, based on always or never using the

EAT or based on arrival time; and (b) dynamic rules that use random forests to assign taxiways on a per aircraft basis.

2 - Is Stochastic Runway Scheduling Computationally

Implementable in Practice?

Senay Solak, Assistant Professor, University of Massachusetts,

Isenberg School of Management, Amherst, MA, 01003, United

States of America, [email protected], Gustaf Solveling,

John-Paul Clarke, Ellis Johnson

We consider a stochastic version of the runway scheduling problem, and analyze it by developing alternative formulations and solution approaches. The results from the computational tests on realistic data sets indicate that practically implementable truncated versions of the proposed solution methods almost always produce high quality solutions.

3 - A Taxiway Planning Model Considering Runway Exit Selection for Landing Aircrafts

Wenda Liu, Tsinghua University, Main Building 616, Beijing,

China, [email protected], Mo Yang, Peng Cheng

The efficiency of taxi process restricts the full use of airport runway capacity. We investigate the taxiway planning problem and discuss the significance of runway exit selection for landing aircrafts. A model based on mixed integer programming is proposed for the assignment of taxi routes and times. We test the model performance with taxiway operations of Beijing Capital International Airport.

About 6.5 percent of taxi time can be saved, and calculation time is acceptable for application.

TD61

61- Russell- Hyatt

Decision and Risk Analysis for Emergency

Management and Terrorism Risk

Cluster: Applications in Emergency Management and

Terrorism Security

Invited Session

INFORMS Phoenix – 2012

Chair: Gregory Parnell, United States Military Academy, West Point, NY,

10996, United States of America, [email protected]

1 - Analyzing the Global Nuclear Detection Architecture

Gregory Parnell, United States Military Academy, West Point, NY,

10996, United States of America, [email protected],

Jason Merrick

The Global Nuclear Detection Architecture (GNDA) uses a multi-layered, time phased, defense in depth strategy. The architecture uses materials protection, control, and accountability; security of radioactive sources, point of departure screening, at-sea interdiction, border protection, and maritime inspection. We provide the structure of an integrated modeling framework to use for system architecture trade-off analyses.

2 - Cyber Risk to Transportation Sector Industrial Control Systems

Barry Ezell, Chief Scientist, Virginia Modeling, Analysis and

Simulation Center, 1030 University Blvd, Suffolk, VA, 23435,

United States of America, [email protected], Mike Robinson,

Joe Weiss

The presentation describes an bridge tunnel and traffic operations center scenario to assess the impact of cyber intrusion in a surface transportation system. PRA simulation estimates the likelihood of each scenario. The cyber effects are injected into regional transportation simulation to assess the consequences of the scenarios. The desired outcome is to raise awareness of cyber issues associated with ICS and expose knowledge gaps and cyber vulnerability to critical infrastructure.

3 - Conceptualizations of Terrorism Risk

Seth Guikema, Assistant Professor, Johns Hopkins University, 313

Ames Hall, Department of Geog & Env. Engineering, Baltimore,

MD, 21218, United States of America, [email protected],

Terje Aven

There are many different conceptualizations of terrorism risk in use, and there is not yet agreement on a coherent framework for addressing terrorism risk. In this talk we give an overview of different ways of defining terrorism risk and the techniques used as measures of terrorism risk. We posit a framework that can help bring clarity to terrorism risk assessments.

4 - Strategic Planning for Nlets – A MODA “Soft-Skills” Approach

Terry Bresnick, Senior Principal Analyst, Innovative Decisions, Inc,

Vienna, VA, United States of America, [email protected], Steve Connell

For the last decade, Nlets, the International Justice and Public Safety Information

Sharing Network, has used a facilitated MODA process to develop, update, and monitor its strategic plan. Nlets uses a Value-Focused Thinking approach to its strategic goals that includes operating in a high-quality, secure environment; growing the membership base; ensuring adequate staffing; ensuring adequate revenue; and reinvesting in its membership community.

TD62

62- Borein A- Hyatt

Agents and Machine Learning Techniques for Auctions

Cluster: Auctions

Invited Session

Chair: Wolf Ketter, Associate Professor of Information Systems,

Erasmus University, Rotterdam School of Management, Rotterdam,

Netherlands, [email protected]

1 - Structural Analysis and its Application to Multi-unit Sequential

Dutch Auctions

Yixin Lu, PhD Candidate, Erasmus University, Rotterdam School of

Management, Rotterdam, Netherlands, [email protected], Eric van Heck,

Alok Gupta, Wolf Ketter

We develop a structural model for multi-unit sequential Dutch auctions where multiple units of identical products are auctioned in sequential rounds. Using the real-world transaction data from a complex Business-to-Business (B2B) auction market, we demonstrate the effectiveness of our model in characterizing the dynamics of winning bids.

2 - Optimal Allocation for Display Advertising

De Liu, Associate Professor, Gatton College of Business and

Economics, University of Kentucky, Lexington, KY, United States of America, [email protected], Huaxia Rui, Andrew Whinston

The spread of real-time display advertising creates a new challenge: how to efficiently allocate countless categories of impressions. We propose a new contingent contract based approach, in which a provider can offer each ad buyer variable quantities contingent on the realized number of impressions. The key advantages of this approach include accommodating heterogenious risk preferences, achieving better allocation efficiency, and joint allocation of multiple categories.

3 - A Kernel-based Combinatorial Auction

Sebastien Lahaie, Microsoft Research,

1290 Avenue of the Americas, New York, NY,

United States of America, [email protected]

We present an iterative combinatorial auction that offers modularity in the choice of price structure, drawing on ideas from kernel methods. The auction is able to detect whether price complexity must be increased to clear the market, and converges to a sparse representation of nonlinear clearing prices. We also show that regularization enables the auction to compute approximate truth-inducing payments in a single run, in contrast to VCG payments which require as many runs as there are bidders.

4 - Approximating Equilibria in Sequential Auctions

Eric Sodomka, PhD Candidate, Brown University, 115 Waterman

Street, 4th Floor, Providence, RI, United States of America, [email protected], Amy Greenwald, Jiacui Li

We search for symmetric equilibria in sequential auctions by modeling the bidder’s problem as a Markov decision process (MDP) and searching for stable transition probabilities, given bidder MDP policies. We quantify how approximately optimizing with respect to approximately stable transition probabilities corresponds to approximate equilibria. We converge to known equilibria for simpler problems and find new approximate equilibria for problems where analytical solutions are unknown.

360

INFORMS Phoenix – 2012

TD63

63- Borein B- Hyatt

Behavioral Theory in Operations Management

Sponsor: Behavioral Operations

Sponsored Session

Chair: Andrew Davis, Assistant Professor of Operations Management,

Cornell University, 401J Sage Hall, Ithaca, NY, 14853,

United States of America, [email protected]

1 - A Tale of Two Countries: Trust and Trustworthiness in China and the U.S.

Yanchong Karen Zheng, Massachusetts Institute of Technology,

Sloan School of Management, 100 Main Street, Cambridge, MA,

United States of America, ya[email protected], Ozalp Ozer,

Yufei Ren

We experimentally investigate how cultural distinctions between China and the

U.S. affect the efficacy of forecast sharing in a supply chain. We demonstrate that trust and trustworthiness are lower and decline more evidently with increasing risks in China. In addition, repeated interactions lead to reversed dynamics for the evolution of trust and trustworthiness in the two countries.

2 - A Laboratory Comparison of Auctions and

Sequential Mechanisms

Anthony Kwasnica, Associate Professor of Business Economics,

Pennsylvania State University, 332 Business Building, University

Park, PA, 16802, United States of America, [email protected],

Andrew Davis, Elena Katok

When bidders incur a cost to learn their valuations, two common selling mechanisms are an English auction, and a sequential bidding process. We experimentally compare the two mechanisms, varying the entry cost, and find that, contrary to theory, average seller revenues are higher under the sequential mechanism, while average bidder profits are the same. We develop a model of noisy bidder entry costs that organizes the experimental data well.

3 - Learning and Forgetting: Vendor Quality Improvement

Anupam Agrawal, University of Illinois at Urbana-Champaign,

363 Wohlers Hall, Champaign, IL, 61820, United States of

America, [email protected], Suresh Muthulingam

We assess the relative benefits of quality improvement initiatives (process improvement, quality assurance, and design improvements) and compare the learning and forgetting from these initiatives to those in autonomous learning.

Our analysis of 4,700 initiatives indicates that (i)Induced learning decays at a slower rate than autonomous learning (ii)Quality Assurance initiatives provide more stable quality improvement results: the benefits persist over time.

4 - Reference Prices and Transaction Utility in Inventory Decisions

Jordan Tong, Assistant Professor, University of Wisconsin-Madison,

975 University Avenue, Madison, WI, 53706, United States of

America, [email protected], Jing-Sheng Song

We present a descriptive model of the effects of reference prices and transaction utility in a newsvendor setting. Our model predicts that an individual’s order is irrationally increasing in past purchasing costs, decreasing in past selling prices, and decreasing in the proportion of high profit margin to low profit margin products in the portfolio. Three laboratory experiments support the model’s predictions. Finally, we analytically investigate the implications of the model to a supply chain.

TD66

66- Ellis West- Hyatt

Joint Session DM/QSR: Various Applications in Text

Mining

Sponsor: Data Mining & Quality, Statistics and Reliability

Sponsored Session

Chair: Seoung Bum Kim, Associate Professior, Korea University, Seoul,

Korea, Republic of, [email protected]

1 - Detection of Changes in Blogger Sentiment using SVM-based

Control Charts

Sungim Lee, Associate Professor, Dankook University, Korea,

Republic of, [email protected]

These days hundreds of millions of Internet users show personal emotions, feelings or thoughts through Blog posts on social media services. In this paper, we will show how to detect the changes in the blogger sentiment data using mulitvariate control charts.

TD67

2 - Text Mining Approach to Find the Key Sentences in

Large Documents

Sugon Cho, Korea University, Seoul, Korea, Republic of, [email protected], Seoung Bum Kim

Finding the key sentences and paragraphs is an important and challenging problem. In recent years, the amount of text data has grown astronomically and this growth has produced a great demand for text summarization. In the present study we applied the text mining methods to find the key sentences from large documents.

3 - Discovery of Associated Patterns in the U.S. Patent Database using Text Mining Approaches

Jeonghun Kim, Korea University, Seoul, Korea, Republic of, [email protected], Seoung Bum Kim

Patent is one of the largest sources of technological information. In this study, we elicit the associative patterns of patents by using various text mining algorithms.

Real data were used to extract meaningful patterns from patent data.

4 - Hybrid Approach for Recommendation System with Contentbased Features from Online Reviews

Minhoe Hur, Seoul National University, Seoul, Korea, Republic of, [email protected]

Content features are widely used in recommendation system. Most previous works considered them as genre, cast or director. But they were superficial and had limitations to show the nature of contents. In this study, we proposed hybrid approach for recommendation. And the contents similarity was calculated with the extracted features from online reviews of the content after latent semantic analysis had been applied. We compared the performance measure to other methods in the conclusion.

TD67

67- Ellis East- Hyatt

Sensor-based Monitoring, Modeling and Analysis I

Sponsor: Quality, Statistics and Reliability

Sponsored Session

Chair: Hui Yang, Assistant Professor, University of South Florida,

Tampa, FL, 33620, United States of America, [email protected]

1 - Cutter Tilt Modeling and Monitoring in Face Milling using

High-definition Metrology

Hui Wang, University of Michigan, Ann Arbor, MI,

United States of America, [email protected]

This research uses surface data measured by high-definition metrology to monitor machine spindle setup and deflection that impact surface quality and part functions. Procedures were developed to estimate and monitor the hard-tomeasure cutter tilts by integrating cutting force modeling and surface data patterns. A Ford case study has demonstrated the effectiveness of the proposed method. This research also establishes a theoretical framework of surface characterization driven by process physics.

2 - Optimal Sensor Placement for Partially Diagnosable

Multi-station Assembly Processes

Zhenyu (James) Kong, Assistant Professor, Oklahoma State

University, 322 Engineering North, Stillwater, OK, 74078, United

States of America, [email protected], Kaveh Bastani

Sensor placement optimization aims to provide the desired level of diagnosability using minimal number of sensors in multi-station assembly processes. Previously proposed sensor optimization methods cannot handle the under-determined systems in which number of sensors is less than that of the process errors. We propose a sensor placement optimization method based on our newly developed diagnosability criteria which is the minimal mutual coherence to uniquely identify the process faults.

3 - Capability Enhanced Adaptive Sensor Allocation Strategy for

Process Monitoring and Diagnosis in a Bayesian Network

Kaibo Liu, Research Assistant, H. Milton Stewart School of

Industrial and Systems Engineering, 765 Ferst Drive, Room 109,

Georgia Inst