2015 Informs Annual Meeting

MA28

INFORMS Philadelphia – 2015

2 - Using an AHP Approach for Eyewitness Identification Enrique Mu, Carlow University, Pittsburgh, PA, United States of America, emu@carlow.edu, Tingting Rachel Chung, Lawrence Reed

4 - Mechanism Design for Team Formation Yevgeniy Vorobeychik, Vanderbilt University, 401 Bowling Ave, Nashville, TN, United States of America, eug.vorobey@gmail.com, Mason Wright We present the first formal mechanism design framework for team formation, building on recent combinatorial matching market design literature. We exhibit four mechanisms for this problem, two novel, two simple extensions of known mechanisms from other domains. We use extensive experiments to show our second novel mechanism, despite having no theoretical guarantees, empirically achieves good incentive compatibility, welfare, and fairness.

Eyewitnesses of a crime are usually asked to identify a potential criminal out of a lineup of suspects. An online experiment using Amazon MT was conducted. Results show that an AHP approach may offer better eyewitness identification success and more importantly less false positive identification ratios than currently sequential lineup approach. 3 - Modeling the Sensitivity and Stability of Preferences Among Colorectal Cancer Screening Alternatives Magda Gabriela Sava, PhD Candidate, Joseph M. Katz Graduate School of Business, University of Pittsburgh, 241 Mervis Hall, Patients are faced with multiple alternatives when selecting the preferred method for colorectal cancer screening, and there are multiple criteria to be considered in the decision process. We model the patient’s choice using an Analytic Network Model and propose a new approach for characterizing the idiosyncratic preference regions for each patient. We show how to use that characterization to derive insights as to the sensitivity and stability of a patient’s individual choice of alternative. 4 - A Stakeholder-theory Based Employer Health Plan Selection Model Mehdi Amini, Professor, The University of Memphis, Department of Marketing & SCM, Fogelman College of Business & Economics, Organizations are called to re-evaluate current plan offerings and potentially, for the first time, select new healthcare providers and policies to ensure that a minimum level of coverage required by law. A new stakeholder-theory based Analytic Network Process (ANP) model is developed to capture a health plan selection decision with the consideration of multiple stakeholders’ interests. What-if analysis is used to explore the robustness of the selected plan. MA28 28-Room 405, Marriott Matching Markets and Their Applications Cluster: Auctions Invited Session Chair: Thayer Morrill, NC State University, Raleigh, NC, United States of America, thayer_morrill@ncsu.edu 1 - Incentives in the Course Allocation Problem Hoda Atef Yekta, University of Connecticut School of Business, Kominers et al. (2011) introduced a heuristic for comparing incentives among the course allocation problem (CAP) algorithms. We investigate their method and adapt it to a more realistic setting with course overlap and a limited number of courses for each student. We compare algorithms including the bidding-point mechanism, the draft mechanism, and recently proposed algorithms like the proxy-agent second-price algorithm in their vulnerability to non-truthful bidding. 2 - Near-optimal Stochastic Matching with Few Queries John Dickerson, CMU, 9219 Gates-Hillman Center, Pittsburgh, PA, 15213, United States of America, dickerson@cs.cmu.edu, Avrim Blum, Nika Haghtalab, Ariel Procaccia, Tuomas Sandholm, Ankit Sharma In kidney exchange, patients with kidney failure swap donors. Proposed swaps often fail before transplantation. We explore this phenomenon through the lens of stochastic matching, which deals with finding a maximum matching in a graph with unknown edges that are accessed via queries, and its generalization to k-set packing. We provide adaptive and non-adaptive algorithms that perform very few queries, and show that they perform well in theory and on data from the UNOS nationwide kidney exchange. 3 - The Secure Boston Mechanism Thayer Morrill, NC State University, Raleigh, NC, United States of America, thayer_morrill@ncsu.edu, Unut Dur, Robert Hammond We introduce the first mechanism that Pareto dominates the Deferred Acceptance algorithm (DA) in equilibrium. Our algorithm, the Secure Boston Mechanism (sBM), is a hybrid between the Boston Mechanism and DA. It protects students that are initially guaranteed a school but otherwise adjusts priorities based on student rankings. We demonstrate that sBM always has an equilibrium that weakly dominates the DA assignment, and that in equilibrium no student receives worse than a fair assignment. Pittsburgh, PA, 15260, United States of America, mgsava@katz.pitt.edu, Luis Vargas, James G. Dolan, Jerrold H. May Memphis, TN, 38152, United States of America, mamini@memphis.edu, Orrin Cooper, Mike Racer Storrs, CT, United States of America, Hoda.AtefYekta@business.uconn.edu

MA29 29-Room 406, Marriott Applied Analytics Across Industries

Sponsor: Analytics Sponsored Session

Chair: Polly Mitchell-Guthrie, Sr. Mgr., Advanced Analytics Customer Liaison Group, SAS Institute, SAS Campus Dr., Cary, NC, 27513, United States of America, Polly.Mitchell-Guthrie@sas.com 1 - Tracking the Regional Economy in Real Time (through Rain and Snow)

Michael Boldin, Federal Reserve Bank of Philadelphia, 10 Independence Mall, Philadelphia, PA, 19106-1574, United States of America, Michael.Boldin@phil.frb.org

The project involves enhancing real-time econometric tracking models for a regional economy to use weather station measurements. Most econometric models use pre-filtered data that excludes seasonal patterns that can distort the effects of important weather events. This project makes use of data that is not pre- filtered and simultaneously derives normal seasonal patterns, the effects of specific weather events, and a measure of the adjusted ‘health’ of the regional economy. 2 - Threadlab: An Analytics Driven Online Clothing Service for Men ThreadLab is a startup company that provides a convenient and customer-friendly online clothing service to men. It elegantly solves a common challenge for a majority of men: Men simply do not like to shop for clothes. ThreadLab takes the work out of clothes shopping by moving the entire decision process onto an analytics platform. All decisions at ThreadLab (from what to stock, what to ship, etc.) are driven by analytical techniques such as mathematical modeling and optimization. 3 - Geospatial Analysis of Bike Share Data Matthew Windham, Director, Analytics, NTELX, Inc., 1945 Old Gallows Rd, Vienna, VA, 22182, United States of America, mwindham@ntelx.com We will explore an end-to-end example of processing Washington DC Bike Share data with BASE SAS. We will walk through the data ingest, cleaning, analysis, and visualization. The results will be visualized in Google Earth. All of the SAS code will be made available to attendees, including the code to write Google Earth KML files that underpin the visualization and exploration capabilities. MA30 30-Room 407, Marriott 2015 Edelman Finalists Reprise Sponsor: CPMS Sponsored Session Chair: Pooja Dewan, BNSF Railway, Fort Worth, TX, 76092, United States of America, Pooja.Dewan@bnsf.com 1 - Maximizing U.S. Army’s Future Contribution to Global Security using Capability Portfolio Analysis Matthew Hoffman, Sandia National Laboratories, P.O. Box 5800 MS 1188, Albuquerque, NM, 87185-1188, United States of America, mjhoffm@sandia.gov, Scott Davis, Shatiel Edwards, John Toczek, ThreadLab, Philadelphia, PA, United States of America, toczek@gmail.com

David Bassett, Gerald Teper, Brian Alford, Craig Lawton, Liliana Shelton, Stephen Henry, Darryl Melander, Frank Muldoon, Roy Rice, Michael McCarthy, Scott Johnson

The Army and supporting team developed and applied the Capability Portfolio Analysis Tool (CPAT), which employs a novel multi-phase mixed integer linear program to optimize fleet modernization problems under complex cost, production, and schedule constraints. Army leadership can now base investment decisions on rigorous portfolio analytics, allowing billions of taxpayer dollars to be optimally prioritized and providing maximum capability and protection to U.S. troops in the decades to come.

154