Informs Annual Meeting 2017

SD49

INFORMS Houston – 2017

SD49

describe our approach, the key design decisions of the system, and our interactions with the city of Los Angeles and commercial entities around fielding the technology. 4 - Personalized Menu Optimization for an Incentivized Smart Mobility Solution

361B What is Behavioral Aspects of OR? Invited: InvitedBehavioral Aspects of OR Invited Session

Bilge Atasoy, MIT, 77 Massachusetts Avenue, MIT.1-180, Cambridge, MA, 02139, United States, batasoy@mit.edu, Xiang Song, Mazen Danaf, Moshe Ben-Akiva

Chair: Martin Kunc, University of Warwick, University of Warwick, Coventry, CV4 7AL, United Kingdom, martin.kunc@wbs.ac.uk 1 - Moderator Martin Kunc, University of Warwick, WBS.Social Studies Building, Office E1.07, Coventry, CV4 7AL, United Kingdom, martin.kunc@wbs.ac.uk The panel will discuss three aspects that differentiate behavioral operations research (OR) from behavioral operations management. Firstly, “content” oriented Behavioral OR. Secondly, “process” oriented Behavioral OR. Thirdly, the “effect” of behavioral and cognitive factors on the activity of analysis and modelling itself. The panel will also discuss how these aspects can form a bridge between both perspectives. 2 - Panelist Leroy White, University of Bristol, School of Eonomics Finance and Accounting, 12 Priory Road, Bristol, BS8 1TU, United Kingdom, leroy.white@bristol.ac.uk 3 - Panelist Sally C. Brailsford, University of Southampton, School of Management, Southampton, S017 1BJ, United Kingdom, s.c.brailsford@soton.ac.uk 4 - Panelist Oliva Rogelio, Texas A&M.University, Mays Business School, College Station, TX, United States, roliva@mays.tamu.edu 361C Sustainable Travel Incentives Sponsored: TSL, Urban Transportation Sponsored Session Chair: Bilge Atasoy, MIT, MIT, Cambridge, MA, 02139, United States, batasoy@mit.edu 1 - Reward Allocation for Maximizing Energy Saving in a Transportation System Adewale Oduwole, University of Massachussets, Amherst, MA, 01002, United States, aoduwole@umass.edu, Song Gao, Hyoshin Park, Moshe Ben-Akiava We study the optimal reward allocation problem to maximize energy saving in a transportation system. In the first version, no personalized menu optimization is considered. In the second version, a bi-level optimization problem is formulated where the high level optimization aims at maximizing system-wide energy saving, while a low-level consumer surplus maximization problem is solved for each user. Numerical tests are carried out study the effects of an array of system parameters, e.g., token budget and menu length, on the solutions. An array of parameterized token allocation strategies are proposed, and their performances against the optimal solution are investigated. 2 - Developing a Multi-scale and Multi-modal Agent-based Modeling SD50 Hesham Rakha, Professor of Civil and Environmental Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA, 24061, United States, hrakha@vt.edu This presentation will describe the various building blocks of the PARC and VTTI’s proposed modeling tool, that include: (a) traveler behavior modeling; (b) modeling of train dynamics; (c) agent-based modeling of driver and vehicle longitudinal and lateral movement; (d) modeling of multi-modal energy/fuel consumption (light duty gasoline vehicles, electric vehicles, plugin hybrid electric vehicles, trucks, Compressed Natural Gas (CNG) buses, and electric trains); (e) integration of microscopic and mesoscopic traffic modeling into a fully-integrated multi-scale multi-modal agent-based modeler. 3 - The COPTER System: An Overview and Preliminary Testing in the Greater LA Area Matthew Klenk, PARC, 3333 Coyote Hill Rd, Palo Alto, CA, 94304, United States, Matthew.Klenk@parc.com COPTER identifies the energy-efficient routes most likely to be adopted by a traveler. The COPTER project brings together expertise in behavioral modeling and analytics algorithms to understand travel needs of users, their constraints, and how likely they are to respond to suggested travel options. In this talk, we Framework for Assessing Network-wide Energy/Fuel Consumption Impacts of Travel Incentive Strategies

We present a personalized menu optimization model for an app-based travel incentive system, Tripod. For each traveler arriving to the system, a customized set of options are presented based on the preferences of the traveler in real-time. The preferences are updated as the traveler performs new trips through an MCMC procedure. Therefore, the personalized menu is based on the most up to date behavioral update. Incentives are allocated based on the guidance from a system optimization model that minimizes network wide energy consumption. Different formulations are evaluated under different scenarios in order to show the added value of personalized nature of the model and the updated preferences. 5 - incenTrip: A Real-Time Personalized Information and Incentivization Tool as a Smart-Mobility Solution Chenfeng Xiong, University of Maryland-College Park, 149 Westway, Apartment 201, Greenbelt, MD, 20770, United States, cxiong@umd.edu, Lei Zhang Different types of incentives in the transportation systems have recently drawn increasing attention to form smart-mobility solutions. Instead of using presumed and fixed amount of incentives, we develop an integrated, personalized and real- time traveler information and incentive technology, incenTrip, to incentivize more energy efficient travel, such as eco-driving, flex-time departures, and ride- sharing. Available in Android and iOS systems, incenTrip smartphone app provides precise and predictive traffic information, coupled with personalized and pro-social incentives, to help its users make smarter travel decisions. Driven by a computationally efficient agent-based modeling system modeling, incenTrip technology simulates the entire transportation systems in real-time and employs artificial-intelligence algorithms to predict the traffic evolution. This online modeling leverages real-time transportation big-data streams supplied by one of the biggest transportation data warehouses in the U.S., the Regional Integrated Transportation Information System (RITIS), hosted at the University of Maryland. Then a control optimization system is developed based on behavior research, travel intent prediction, and incentive optimization, in order to provide optimized incentives and route/departure time guidance. Chair: Yan Huang, University of Michigan, yphuang@umich.edu 1 - Data-driven Promotion Planning for Mobile Applications Manqi Li, University of Michigan, 710 Tappan Avenue, Ann Arbor, MI, 48105, United States, manqili@umich.edu, Yan Huang, Amitabh Sinha The mobile app market is one of the fastest growing markets. In this work, we propose a two-step analytical approach to optimize the price promotion strategy for mobile apps. In the first step, we empirically estimate the immediate and long- term effects of price promotions on app download volume. The estimation results reveal two interesting characteristics: (1) promotion effect; and (2)visibility effect. Based on the estimated download function, we then formulate a longest-path problem to solve for optimal policy. The resulting optimal promotion strategy makes the best use of visibility effect and significantly increases developers’ long- term revenue compared with their current practice. 2 - Linking Clicks to Bricks: Estimating Causal Effects of Email Ads on Offline Sales Mi Zhou, Carnegie Mellon University, 4800 Forbes Ave, Pittsburgh, PA, 15213, United States, mzhou1@andrew.cmu.edu, Vibhanshu Abhishek, Ritwik Sinha, Nikos Vlassis Online advertising market is soaring and has attracted the attention of many researchers. However, research in this area has almost entirely focused on consumers’ online behaviors and has not incorporated their decision-making in the offline setting, which may lead to incomplete understanding of online advertising effectiveness in today’s multi-channel world. In this paper, we use a Big Data which links consumers’ online behaviors to their purchase records offline, and find that email ads have non-linear positive effects on offline sales. Further, we use a supervised Machine Learning method to estimate heterogeneous effects of email ads in different subpopulations partitioned by causal trees. SD51 361D Digital Markets and Platforms Sponsored: Information Systems Sponsored Session

118