2015 Informs Annual Meeting

MB65

INFORMS Philadelphia – 2015

MB65 65-Room 113B, CC Risk Attitudes in Decision Analysis Sponsor: Decision Analysis Sponsored Session

2 - Similar Days? A Story Based on User-defined Similarity Yi Liu, UC Berkeley, 107 McLaughlin Hall, Berkeley, CA, 94720, United States of America, liuyi.feier@gmail.com, Mark Hansen, Alexey Pozdnukhov In this work, we propose a supervised data-mining algorithm for measuring similarity between two days. First, the algorithm trains the distance matrix between hours according to user-defined similarity and dissimilarity. Then it calculates the daily distance as a weighted sum of hourly distances. The approach can be applied to measure similarity between two days post-operation or identify similar days in the past for a given day. 3 - Representative Traffic Management Initiative Decisions Alex Estes, University of Maryland-College Park, 3117 AV Williams, College Park, MD, 20742, United States of America, aestes@math.umd.edu, Michael Ball, David Lovell We provide a method for presenting data on traffic management initiatives so that it may more easily be interpreted by researchers or by TMI decision makers. This method involves solving a dominating set problem to produce a set of TMIs which are representative of the range of TMI decisions that have been taken in the past. MB67 67-Room 201A, CC The Role of Information in Transportation Models Sponsor: TSL/Freight Transportation & Logistics Sponsored Session Chair: Dirk Mattfeld, Germany, d.mattfeld@tu-braunschweig.de 1 - Approximate Dynamic Programming for Temporal-spatial Anticipation and Routing of Service Requests Justin Goodson, Saint Louis University, St. Louis, MO, United States of America, goodson@slu.edu, Marlin Ulmer, Dirk Mattfeld We consider the problem of dynamically routing a vehicle to respond to service requests arriving randomly over a given time horizon. Building on a value- function approximation (VFA) that estimates rewards-to-go via the temporal components of the state variable, we use rollout algorithms to explicitly consider spatial dimensions. Our method improves upon the VFA scheme in isolation and points to the potential benefit of using two different approximate dynamic programming techniques in tandem. 2 - A Look-ahead Solution Framework for the Dynamic Vehicle Routing Problem Han Zou, University of Southern California, Los Angeles, CA United States of America, hanzou@usc.edu, Maged Dessouky The problem concerns routing a fleet of capacitated vehicles in real time to fulfill orders placed by a known set of customers. Some of the orders are known a priori, while the rest are placed in real time. We develop a look-ahead solution framework that periodically re-optimizes current vehicle routes by using both realized and forecasted information. With fine-tuned parameter settings, our approach has the potential to outperform both a priori routing and total dynamic dispatching schemes. 3 - Optimal Information Collection in the Vehicle Allocation Problem in Mega-Cities Yixiao Huang, Tsinghua University, China, huangyx12@mails.tsinghua.edu.cn, Warren Powell, Ilya Ryzhov, Lei Zhao In mega-cities, urban freight delivery companies serve customers spread in a large area. These companies typically divide the city into several regions and allocate the delivery vehicles to serve these regions. The precise cost evaluation of a vehicle allocation decision can be very expensive, due to the volume and uncertainty of the delivery demand. We study on, given a fixed information collection budget, how to collect the cost evaluation information (possibly through real-life practice) to gain the best knowledge to allocate the vehicles optimally. 4 - Data-Driven Vehicle Routing with Profits

Chair: Andrea Hupman, University of Missouri-St. Louis, 1 University Blvd, St. Louis, MO, United States of America, hupman1@illinois.edu 1 - Individualized Predictions of Normative Decision Making Andrea Hupman, University of Missouri-St. Louis, 1 University Blvd, St. Louis, MO, United States of America, hupman1@illinois.edu, Ali Abbas Understanding how individuals make decisions in practice and predicting behavior is important in many practical applications, giving rise to numerous descriptive models of decision making behavior. In this talk, a behavioral decision making experiment is described. The results are consistent with a bounded rationality explanation of decision making behavior in which making accurate predictions of decisions hinges on individualized information about the decision maker’s risk attitude. 2 - Marriage and Managers’ Attitudes to Risk Pavel Savor, Fox School of Business, Temple University, Philadelphiaa, PA, pavel.savor@temple.edu We explore the impact of marriage on corporate CEOs and find that firms run by single CEOs exhibit higher stock return volatility, pursue more aggressive investment policies, and do not respond to changes in idiosyncratic risk. These effects are weaker for older CEOs. Our results continue to hold when we use variation in divorce laws across states to instrument for marital status, supporting the hypothesis that marriage itself drives choices rather than it just reflecting innate heterogeneity. 3 - Using Means Objectives to Present Risk Information Candice Huynh, Cal Poly - Pomona, CA, United States of America, candicehuynh@cpp.edu, Jay Simon When making decisions involving alternatives with risk, individuals are not always able to express or view the possible outcomes in terms of a fundamental objective. To apply information about a means objective correctly, a decision maker must first translate it into information about a fundamental objective. This paper presents the results of a study regarding decision makers’ preferences when information is presented either in terms of a means objective or a fundamental objective. 4 - Evolving Risk Perceptions of Cybersecurity Events Heather Rosoff, Research Assistant Professor, University of Southern California, Price School of Public Policy & CREATE, 3710 McClintock Avenue, Los Angeles, CA, 90089-2902, United States of America, rosoff@usc.edu, Robin Dillon-Merrill Mitigating cyber risks requires understanding how people evaluate risks in this challenging context and in particular, how they respond to repeated warnings over time. In the case of the Target credit card breach in 2013, evidence now shows that Target had failed to follow through on security alerts triggered by the hacker’s activities. We speculate on why Target’s security team, and why people in general, might not respond to such alarms in the cyber-security context. Learning from History Sponsor: Aviation Applications Sponsored Session Chair: Yi Liu, UC Berkeley, 107 McLaughlin Hall, Berkeley, CA, 94720, United States of America, liuyi.feier@gmail.com 1 - The Identification of Similar Days in the New York Area for Air Traffic Flow Management Initiatives Kenneth Kuhn, RAND Corporation, 1776 Main Street, Santa Monica, CA, 90407, United States of America, kkuhn@rand.org, Akhil Shah Analysis of air traffic flow management initiatives can show the relative success of decisions, but must account for conditions during planning and operations. We apply cluster analysis to identify similar days, using features detailing aviation weather and air traffic around New York. An example is the degree to which jet route J75 is blocked by convective weather at 9am according to a 7am weather forecast. Some features are based on automated approaches such as Principal Component Analysis. MB66 66-Room 113C, CC Air Traffic Management Decision Support:

Dauwe Vercamer, Ghent University, Ghent, Belgium, Dauwe.Vercamer@ugent.be, , Dirk Van Den Poel, Michel Gendreau, Philippe Baecke

In sales teams, making optimal visits is crucial. Customer Profitability models help in identifying top customers, but do not consider the associated visit costs. Vehicle Routing models make efficient schedules, but use naive estimates rather than good forecasts. Our prescriptive analytics approach uses auxiliary data and statistical learning to approximate full-information vehicle routes. The results show this approach improves customer selection in vehicle routes and maximizes profitability.

194