2015 Informs Annual Meeting

SC27

INFORMS Philadelphia – 2015

2 - The Value of Fit Information in Online Retail Antonio Moreno-Garcia, Northwestern University, 2001 Sheridan Rd, Evanston, Il, 60208, United States of America, a-morenogarcia@kellogg.northwestern.edu, Santiago Gallino We conduct a field experiment to quantify the value of fit information in online retail. 3 - The Effect of Music Labels on Song Popularity in Electronic Markets Without Barriers-to-Entry Marios Kokkodis, Assistant Professor, Boston College, 34 E 10th, New York, NY, 10009, United States of America, mkokkodi@stern.nyu.edu In this work we study the effect of music labels on song popularity in electronic markets without barriers to entry. 4 - The Moderating Effects of Product Attributes and Reviews on Recommender System Performance Dokyun Lee, Carnegie Mellon University, United States of America, leedokyun@gmail.com, Kartik Hosanagar We investigate the moderating effect of several product attributes on the efficacy of a recommender system for increasing conversion rate via a field experiment on one of the top North American retailer’s website. Our results provide many managerial implications on when to utilize recommenders and how recommenders interact with reviews and product attributes. SC26 26-Room 403, Marriott INFORMS Undergraduate Operations Research Prize II Cluster: INFORMS Undergraduate Operations Research Prize Invited Session Chair: Aurelie Thiele, Lehigh University, 200 W Packer Ave, Bethlehem, PA, 18015, United States of America, aut204@lehigh.edu 1 - Piecewise Static Policies for Two-stage Adjustable Robust Optimization Problems under Uncertainty Omar El Housni, Industrial Engineering and Operations Research, We consider two-stage adjustable robust linear optimization problems under uncertain constraints and study the performance of piecewise static policies. We show that surprisingly there is no piecewise static policy with polynomial number of pieces with performance significantly better than a static policy in general. We also present a family of piecewise static policy with exponential pieces that has a significantly better performance than a static solution and admits a compact MIP formulation. 2 - Multi-step Bayesian Optimization for One-dimensional Feasibility Determination Massey Cashore, University of Waterloo, 200 University Ave W, Waterloo, On, N2L 3G1, Canada, masseycashore@gmail.com, Peter Frazier Bayesian optimization methods allocate limited sampling budgets to maximize expensive-to-evaluate functions. One-step-lookahead policies are often used, but computing optimal multi-step-lookahead policies remains a challenge. We consider a specialized Bayesian optimization problem: finding the superlevel set of an expensive one-dimensional function, with a Markov process prior. We compute the Bayes-optimal sampling policy efficiently, and characterize the suboptimality of one-step lookahead. 3 - Alleviating Competitive Imbalances in NFL Schedules: An Integer Programming Approach Kyle Cunningham, Northeastern University, Healthcare Systems Engineering Institute, Boston, MA, United States of America, cunningham.k@husky.neu.edu, Murat Kurt, Niraj Pandey, Mark Karwan While the NFL uses complex rules in scheduling its games, NFL schedules are not robust in creating a consistent competitive appeal. We propose a two-stage MILP approach to reduce competitive disadvantages in schedules arising from various sources including rest differentials due to bye-weeks and Thursday games, long streaks of road games, and short-week travel. Our results for the 2012-2015 seasons indicate that our approach can substantially improve NFL schedules in various fairness metrics. Columbia University, 547 Riverside Drive Apt. 1B, New York, NY, 10027, United States of America, omar.el-housni@polytechnique.edu, Vineet Goyal

4 - Robust Multi-Objective Clustering Andy Zheng, Northwestern University, 1501 Leavenworth, San Francisco, CA, 94109, United States of America, azheng92@gmail.com We propose a multi-objective method that leverages robust optimization for hierarchical clustering (rMOC). rMOC chooses clusters based on a weighted sum of data-intrinsic objective functions, determining a threshold that is most robust to uncertainties in these weights. We compare this method to the reference methods of $K$-means and Gaussian mixture models. In terms of misassignment rate, rMOC outperforms both other methods on several benchmark datasets. 5 - Optimal Resource Allocation in Breast Cancer Screening with Different Risk Groups

Magdalena Romero, Universidad Adolfo Ibañez, Santiago, Santiago, Chile, maromero@alumnos.uai.cl, Qingxia Kong, Susana Mondschein

This paper investigates how many lives can be saved from breast cancer death through optimal allocation of limited resources in public health. We build a two- stage stochastic dynamic programming model to optimally allocate a limited number of mammograms, among women with different risk levels in breast cancer, which is applied to the case in Chile. We find that simply through dividing women into 3 risk groups, we can save 88 lives per 100,000 women, compared to the current practice in Chile.

SC27 27-Room 404, Marriott

MCDM Methods and Applications Sponsor: Multiple Criteria Decision Making Sponsored Session Chair: Luiz Autran Gomes, Ibmec/RJ. Rio de Janeiro – RJ, Brazil, autran@ibmecrj.br 1 - Multi-criteria Evaluation of Sustainable Manufacturing Jian-bo Yang, Professor Of Decision And System Sciences, Manchester Business School, The University of Manchester, Manchester, M15 6PB, United Kingdom, jian-bo.yang@mbs.ac.uk, Panitas Sureeyatanapas Sustainable manufacturing becomes increasingly important. The paper explains how a manufacturer can maintain its business and operations in long term by combining green manufacturing, corporate social responsibility and green supply chain. It then focuses on discussing how criteria and indicators for evaluating progress to sustainable manufacturing can be established. A case study for evaluating sustainability performance in the sugar manufacturing industry of a developing country is discussed. 2 - Recent Development and Applications of Evidential Reasoning Approach for Decision Making Dong-ling Xu, Professor Of Decision Science And Systems, Manchester Business School, The University of Manchester, Manchester, United Kingdom, ling.xu@mbs.ac.uk, Jian-bo Yang We report the recently discovered relationship between Bayesian inference and the Evidential Reasoning approach for multiple criteria decision making under uncertainty. It is significant because it opens up new research avenues in many fields such as the extension of Bayesian inference with imperfect probability information which may not be fully reliable and the enhancement of evidence and random set theories. A few applications are reported with a focus on extracting evidence from big data. 3 - The Primary Aluminum Industry as a Complex Adaptive System David Olson, Professor, University of Nebraska Lincoln, CBA 256, Lincoln, NE, 68588-0491, United States of America, dolson3@unl.edu Supply chains are critically important elements of global business, involving high levels of interdependence. Supply chains have been suggested to be complex adaptive systems. Supply chains usually emerge rather than result from the purposeful design of a single controlling entity. This paper presents the global primary aluminum industry viewed from the perspective of complex adaptive systems. Unintended consequences of actor decisions in this industry in the past forty years. 4 - MCDM Methods Inspired by Prospect Theory Luiz Autran Gomes, Ibmec/RJ. Rio de Janeiro – RJ, Brazil, autran@ibmecrj.br This paper reviews attempts to develop discrete MCDM methods inspired by Kahneman and Tversky’s prospect theory. After going through the first steps making use of linear prospect theory the essentials of the TODIM method and extensions are presented. The paper closes with outlining how qualitative methods of MCDM such as DEX or Verbal Decision Analysis can be combined with TODIM-based methods in order to approach complex decision making problems.

101