2016 INFORMS Annual Meeting Program

WD15

INFORMS Nashville – 2016

WD16 105A-MCC

3 - A Study on Optimal Locations Of Public Facilities To Maximize Social Benefit” Hyunhong Choi, PhD Student, Seoul National University, 1 Gwanak-ro Gwanak-gu, Seoul, 08826, Korea, Republic of, hongchoi@snu.ac.kr, Misuk Lee, Yoonmo Koo Economic feasibility analysis on the public facility construction is widely being studied. However, studies concerning the optimal location of these facilities has not been fully explored yet. In this study, we estimated consumers’ willingness to pay depending on the distance to non-market goods(i.e., arboretums) by using the contingent valuation method. Then, we utilized nested partitions method, one of the categorical optimization methods, to find optimal locations maximizing social benefit among numerous alternatives. 4 - Location Determination Of A Milk Condensing Plant In Tennessee Given the increasing popularity of local foods and the desire to reduce shipping costs and carbon footprint, Tennessee-based dairy product processors are likely interested in sourcing condensed milk from an in-state milk condensing plant. Based on dairy farmer surveys, distances, and transportation costs, this study uses a mixed integer linear programming model to determine the optimal location for a condensing plant that minimizes transportation costs from farms to the condensing plant and the condensing plant to further processors. WD15 104E-MCC Intelligent Information Systems Sponsored: Artificial Intelligence Sponsored Session Chair: Victor Benjamin, University of Arizona, 1130 E Helen St, Tucson, AZ, 85719, United States, vabenji@email.arizona.edu 1 - Using Big Data And Analytics To Enable Smart Mobility Yun Wang, University of Arizona, Tucson, AZ, 85716, United States, yunw@email.arizona.edu, Faiz Currim, Sudha Ram In this work we introduce a three-layer management system to support smart urban mobility with an emphasis on bus transportation. In Layer-1, we apply novel Big Data techniques to efficiently compute bus travel times and passenger demands using universal data streams. Layer-2 contains two analytic components: network analysis of passenger transit patterns and causal relationship analysis for bus delays. The third layer provides interactive visualization tools for decision support. Our system is developed in cooperation with the city of Fortaleza in Brazil. The use of generally available urban transportation data makes our methodology adaptable and customizable for other cities. 2 - Analyst Language In Quarterly Earnings Calls: Comparing Interactions With Fraudulent And Non-fraudulent Managers Lee Spitzley, University of Arizona, lspitzley@email.arizona.edu Corporate financial fraud damages investors, the public, and the companies involved. Fraudulent managers must convince investment analysts who study the company that what they say is accurate and represents the true state of the business. I will examine the content of analyst utterances when they are interacting with the managers during earnings calls. If analysts suspect abnormalities, they may modify their questioning strategies. This study will test for differences in the topic composition of all analysts on a call, and for differences within analysts who follow both fraudulent and non-fraudulent companies in the same industry. 3 - An Empirical Study Of Venders’ Profit Under Different Mechanisms On Online Crowdsourcing Platforms Xiao Han, Shanghai Jiao Tong University, Shanghai, China, hanxiao@sjtu.edu.cn, Pengzhu Zhang Online crowdsourcing markets not only allow buyers of any size to tap into a large talented pool of workforce but also expand the market reach for vendors. Prior studies on crowdsourcing markets have mainly focused on buyers, and there is a lack of understanding of how vendors survive and evolve in the crowdsourcing markets. This paper aims to fill this gap by taking the perspective of vendors and ask how vendors benifit in crowdsourcing markets under different mechanisms. 4 - The Dark Side Of The Singularity: Can OR/MS Help? John D Little, Massachusetts Institute of Technology, Sloan School of Management, Room E62-534, Cambridge, MA, 02142, United States, jlittle@mit.edu The “Singularity” is the point in time when artificial intelligence (AI) exceeds human intelligence. This may occur by putting AI on computers, by biological creation, or by a mixture of both. Some of the people writing about this or developing advanced AI are Victor Vinge, Ray Kurzweil (the Singularity is Near), Tom Malone, Ben Goertzel and Hugo de Garis. The dark side is that most people in this room will be left far behind. Kurzweil notes that AI develops exponentially, whereas most of us extrapolate linearly. David Mendez, Graduate Student, University of Tennessee, 2621 Morgan Circle, Knoxville, TN, 37996, United States, dmendez@vols.utk.edu

Optimization and Learning in Urban Delivery Sponsored: Optimization, Optimization Under Uncertainty Sponsored Session Chair: Lei Zhao, Tsinghua University, Qinghua West Road, Beijing, 100084, China, lzhao@tsinghua.edu.cn 1 - Robust Inventory Management Under Supply And Demand Uncertainties Jie Chu, McMaster University, Hamilton, ON, Canada, chuj6@mcmaster.ca, Kai Huang We simultaneously consider demand and supply uncertainties in a robust optimization framework. We first consider a single-station case, then we extend to a multi-echelon network case. The resulting robust counterpart of the network case does not maintain the same difficulty as its nominal problem. Nonetheless, we present an approximation and thus it can be solved more efficiently. We demonstrate the effectiveness of proposed models numerically. 2 - Learning In Multi-stage Rollouts Saul Toscano-Palmerin, Cornell University, Ithaca, NY, United States, st684@cornell.edu, Peter Frazier We consider a transportation company choosing routes on which to offer service. Each route has an unknown parametric demand distribution, and we wish to choose routes subject to a budget constraint to maximize total demand. We propose a two-stage optimal learning algorithm, where first we offer service on some routes, and learn from observed demand about demand distributions on other similar routes. Then in a second stage, we offer service on additional routes suggested to be good by the first stage. We demonstrate that this two-stage optimal learning approach captures more demand than a one-stage approach that does not leverage the opportunity to learn. 3 - Optimal Learning In Urban Delivery Resource Allocation Yixiao Huang, Tsinghua University, 530 Shunde Building, Beijing, 100084, China, huangyx12@mails.tsinghua.edu.cn, Lei Zhao, Warren B Powell, Ilya O Ryzhov We study knowledge gradient (KG) based optimal learning methods to optimize the urban delivery resource allocation decisions, when the evaluation of such decisions is expansive. Nonlinear Optimization Algorithms II Sponsored: Optimization, Nonlinear Programming Sponsored Session Chair: Daniel Robinson, Johns Hopkins University, 3400 N. Charles Street, Baltimore, MD, 21218, United States, daniel.p.robinson@gmail.com Co-Chair: Frank Edward Curtis, Lehigh University, 27 Memorial Dr, Bethlehem, PA, 18015, United States, frank.e.curtis@gmail.com Co-Chair: Andreas Waechter, Northwestern University, 2145 Sheridan Road, Evanston, IL, 60208, United States, waechter@iems.northwestern.edu 1 - A Geometry Driven Active-set Method For Elastic-net Minimization Daniel Robinson, Johns Hopkins University, daniel.p.robinson@gmail.com We propose an efficient and provably correct active-set based algorithm for solving the elastic net problem. The proposed algorithm exploits the fact that the nonzero entries of the elastic net solution fall into an oracle region, which we use to define and efficiently update an active set. The proposed update rule leads to an iterative algorithm which is shown to converge to the optimal solution in a finite number of iterations. We present experiments on computer vision datasets that demonstrate the superiority of our method in terms of both clustering accuracy and scalability. WD17 105B-MCC

460