2016 INFORMS Annual Meeting Program

TA87

INFORMS Nashville – 2016

3 - An Analysis Of Menus Of Multi-part Tariffs Ryan Choi, Assistant Professor of Marketing and SCM, Eastern Michigan University, 300 W. Michigan Ave., College of Business, Eastern Michigan University, Ypsilanti, MI, 48197, United States, jchoi20@emich.edu, Taewan Kim We study which characteristics of three-part tariffs (3PTs) generate greater profit than two-part tariffs and examine the optimal values of 3PTs. Under an assumption of consumer heterogeneity and a full extraction of low type segment’s surplus, the seller can extract more of high type surpluses. Literatures argue that offering high type contracts only may be more profitable than keeping the low type paying high information rent. Since the firm can charge greater rent from the high type, offering 3PTs contracts to both high and low segments will be more profitable even though the taste parameter is extremely low, regardless of the proportion of the low type.

3 - Online Decision-making With High-dimensional Covariates Hamsa Sridhar Bastani, Stanford University, 10 Comstock Circle, Apt 304, Stanford, CA, 94305, United States, hsridhar@stanford.edu, Mohsen Bayati Big data has enabled decision-makers to personalize choices based on an individual’s observed characteristics. We formulate this problem as a multi-armed bandit with high-dimensional covariates, and present a new efficient algorithm that provably achieves near-optimal performance. The key step in our analysis is proving convergence of the LASSO estimator despite non-iid data induced by the bandit policy. We evaluate our algorithm using a real patient dataset on warfarin dosing; here, a patient’s optimal dosage depends on her genetic profile and medical records. Our algorithm outperforms existing bandit methods as well as physicians to correctly dose a majority of patients. 4 - An Efficient Algorithm For Dynamic Pricing Using A Graphical Representation Swati Gupta, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States, swatig@mit.edu, Maxime Cohen, Jeremy Kalas, Georgia Perakis We study a multi-period, multi-item pricing problem: maximize the total profit by choosing feasible prices that satisfy various business rules. We develop a graphical model that can solve the problem for small memory. We make no assumption on the structure of the demand. For large memory, we show NP-hardness and approximate general demand functions using the reference price model. We give an approximation to solve the latter, extend it to handle cross-item effects among multiple items using the notion of a virtual reference price. We cluster items into blocks and incorporate global business constraints, and finally validate our results on demand models calibrated by real supermarket data. 5 - Evaluating The First-mover’s Advantage In Announcing Real-time Delay Information Siddharth Prakash Singh, PhD Candidate, Tepper School of Business, Carnegie Mellon University, Tepper School of Business, 5000 Fobes Avenue, Pittsburgh, PA, 15213, United States, sps1@andrew.cmu.edu, Mohammad Delasay, Alan Scheller-Wolf We study queueing models of two comparable service providers competing for market share. The announcer (A) voluntarily provides real-time delay information, for example on a website. For the non-announcer (N), customers are only aware of a periodically updated long-term average delay. Customers make patronage decisions based on available delay information. We investigate how A, as the first-mover in announcing real-time delay information, influences market shares and customer delays. We find that when A is not the higher- capacity provider, she benefits on both market share and delay. However, when A is the higher-capacity provider, announcing may result in lower market share or longer delays. TA89 Broadway C-Omni Advances in Traffic Flow Modeling Sponsored: TSL, Intelligent Transportation Systems (ITS) Sponsored Session Chair: Pitu Mirchandani, Arizona Statew University, P.O. Box 878809, Tempe, AZ, 85287, United States, pitu@asu.edu 1 - A Kalman Filter Approach For Dynamic Calibrationof a Simplified Lower-Order Car Following Model Kerem Demirtas, Arizona State University, 699 S. Mill Ave. Tempe, Brickyard Engineering 553, Tempe, AZ, 85281, United States, kdemirta@asu.edu, Pitu Mirchandani, Xuesong Zhou In this study, we are interested in dynamic calibration of car following parameters in order to explore both inter-driver and intra-driver heterogeneity. Specifically, we offer an augmented state space system for a lower order linear spacing car following model developed by Newell and implement a modified Kalman filter algorithm in order to track the leader-follower pairs and simultaneously predict and estimate the parameters related with the behavior of the following drivers. The algorithm is tested on the trajectories from NGSIM data and show satisfactory results. Interpretation of the results and promising future research directions are given. 2 - Efficient Supply Calibration Of Large-ScaleTraffic Simulators Kevin Zhang, Massachusetts Institute of Technology, Cambridge, MA, United States, kzhang81@mit.edu In this presentation, we propose a simulation-based optimization algorithm for the supply calibration of stochastic traffic simulators. We present a metamodel that combines information from the simulator with a problem-specific analytical network model. This metamodel is embedded within a derivative-free trust region algorithm. With this method, we aim to identify transportation-relevant solutions with improved performance within a strict computational budget. The approach is validated on a real traffic network; results are presented

TA87 Broadway A-Omni Panel: Guide to the Analytics Body of Knowledge (ABOK)

Sponsored: Analytics Sponsored Session

Moderator: Louise Wehrle, INFORMS, 5521 Research Park Drive, Catonsville, MD, 21228, United States, louise.wehrle@informs.org 1 - Guideto The Analytics Body Of Knowledge (ABOK) Louise Wehrle, INFORMS, 5521 Research Park Drive, Catonsville, MD, 21228, United States, louise.wehrle@informs.org The Guide to the Analytics Body of Knowledge (ABOK) is being created in support of the Certified Analytics Professional (CAP®) program. The ABOK will serve as a central repository for key analytics knowledge, supported by in-depth subject matter expert interviews and writing as well as formerly unpublished case studies. Join us to learn more about the Analytics BOK and provide your input to the creators of this first edition of the ABOK. 2 - Panelist Terry Harrison, Pennsylvanis State University, University Park, PA, 16802, United States, tharrison@psu.edu 3 - Panelist James Cochran, University of Alabama, Culverhouse College of Commerce & Bus Admin, Tuscaloosa, AL, 35487, United States, jcochran@cba.ua.edu TA88 Broadway B-Omni Service Science Best Student Paper Competition I Award Session Chair: Robin Qiu, Penn State University, 30 E. Swedesford Road, Malvern, PA, 19355, United States, robinqiu@psu.edu 1 - Bike-share Systems: Accessibility And Availability Ashish Kabra, INSEAD, Boulevard de Constance, Fontainebleau, 77305, France, ashish.kabra@insead.edu, Elena Belavina, Karan Girotra This paper estimates the relationship between ridership of a bike-share system and its design aspects— station accessibility and bike-availability. Our analysis is based on a structural demand model that considers the random-utility maximizing choices of spatially distributed users, and it is estimated using high- frequency system-use data from the bike-share system in Paris and highly granular data on sources of bike-share demand. A novel model separates the long-term and short-term effects of higher bike-availability. Because the scale of our data render traditional numerical estimation techniques infeasible, we develop a novel transformation of our estimation problem. 2 - Queues With Redundancy: Is Waiting In Multiple Lines Fair? Leela Aarthy Nageswaran, Carnegie Mellon University, 308 GSIA, Tepper School of Business, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, leelaaarthy@gmail.com, Alan Scheller-Wolf We study the performance of two queues serving two classes of customers, one of which is redundant: a redundant customer joins both queues simultaneously and is “served” when any one of her copies completes service. Applications of redundancy range from supermarkets with multiple checkout lines to multiple listing for kidney transplants. By analyzing different policies that a non-redundant customer may use to join a queue when faced with different levels of system information, our model provides fundamental insights on optimal queue-joining policies and on fairness in such systems.

262