Informs Annual Meeting 2017

MC30

INFORMS Houston – 2017

MC30

4 - The Effects of Thickness and Competition in Ride-sharing Markets Afshin Nikzad, Stanford University, 741A Homer Ave, Stanford, CA, 94301, United States, afshin.nikzad@gmail.com We study the effects of thickness and competition in ride-sharing markets. We vary thickness by varying the size of labor supply while holding the customer side fixed. When the market is sufficiently thick, wage and workers’ average welfare decrease with labor supply. Otherwise, wage and workers’ average welfare increase with labor supply, countering the prediction by the law of demand. Intuitively, workers are “complements” in a thin market -for example, their hourly wage and average employment time goes up with the labor supply- whereas they become “substitutes” and compete with each other in a thick market. 350E Robust Optimization and Machine Learning Sponsored: Artificial Intelligence Sponsored Session Chair: Marek Petrik, University of New Hampshire, 33 Academic Way, Kingsbury W233, Durham, NH, 03861, United States, mpetrik@cs.unh.edu 1 - Efficient Algorithms for Robust Markov Decision Processes with State Rectangularity Clint Chin Pang Ho, Imperial College Business School, London, United Kingdom, c.ho12@imperial.ac.uk, Marek Petrik, Wolfram Wiesemann Robust Markov decision processes (MDPs) seek for optimal policies in view of the worst transition kernel from within an ambiguity set that specifies the knowledge about the unknown true Markov process. Although robust MDPs have emerged as powerful modeling tools, robust MDPs have typically been considered to be intractable, except for special cases where the ambiguity sets are rectangular in both the states and the actions. In this talk, we develop tractable solution techniques for robust MDPs whose ambiguity sets are only required to be rectangular in the states. 2 - Information Collection Optimization in Designing Marketing Campaigns for Market Entry Somayeh Moazeni, Assistant Professor, Stevens Institute of Technology, Babbio Center, 1 Castle Point Terrace on Hudson, Hoboken, NJ, 07030, United States, smoazeni@stevens.edu Developing marketing strategies for market entry is challenging due to the scarcity of relevant data. A sequential information collection optimization assists in creating new data points, given a finite learning budget to identify effective features maximizing the final outcome. We model the marketing campaign performance by a multiplicative advertising exposure model with Poisson jumps whose intensity depends on the marketing features. We develop an optimal learning policy and propose a computationally efficient algorithm to handle the large number of features. The performance of the optimal learning policy is compared to commonly used benchmark policies. 3 - Robust Reinforcement Learning with a Baseline Yinlam Chow, IBM.Research, 1101 Kitchawan Rd, Yorktown, NY, 10598, United States, na, Marek Petrik, Marek Petrik, Marek Ghavamzadeh A fundamental problem in sequential decision-making under uncertainty is to compute a safe policy, i.e., a policy that is guaranteed to have a better performance than a baseline strategy, given a batch of data. In this talk, We describe a model-based approach to this problem, in which the goal is to compute a safe policy, given an inaccurate model of the system with known accuracy guarantees. The inaccurate model and error bound may be constructed using the batch of data and prior knowledge about the system. MC29

350F Supporting Personal Decisions with Social Media Analytics Invited: Social Media Analytics Invited Session Chair: Xue Tan, University of Washington, University of Washington, Seattle, WA, 98105, United States, xuetan@uw.edu 1 - Assessing the Impact of Disclosure Policy on Investment Pattern: Evidence from a Natural Experiment in Supply Chain Finance Market Zhijin.Zhou, University of Washington, Foster School of Business, Seattle, WA, 98195, United States, zjzhou@uw.edu We examine the effect of exogenous information disclosure policy change on projects’ fundraising performances in an online supply chain finance market. Regression Discontinuity analysis was conducted to verify a discontinuous jump in fundraising performances around the shock, Panel model was then used to assess the inside mechanism of the jumps. Results show that the policy change has a positive effect on projects’ fundraising speed and investment rate, while has a negative influence on projects’ average investment amount. Moreover, the influence will be attenuated by the fundraising target and debtors’ ex experience, and strengthen by creditors’ ex experience. 2 - Value of Information in Online Dating: An Empirical Study Kyungmin Park, University of Washington, Seattle, WA, United States, parkk30@uw.edu Online dating service has become a popular source to meet new people. A typical challenge to a firm in this market is matching people who are likely to be attractive to each other. In this regard, improving matching efficiency is important to enhance a firm’s competence in the market. An user wants to find a person not only whom he like, but also who would like him. In this setting, a firm should also consider dyad-level characteristics of each pair of users in addition to individual-level characteristic of each user. This research explore this point by analyzing a rich data from a Korean major online dating application. 3 - Uber Introduction and Spillover Effects on Transportation System: Empirical Study Kyungsun Rhee, University of Washington, 4725 24th Avenue NE, # 405, Seattle, WA, 98105, United States, ksr22@uw.edu, Jinyang Zheng, Yong Tan, Youwei Wang This study aims to investigate whether the entry of Uber influences on the demand patterns in overall transportation system. Using a national panel data set constructed from four sources, we examine the longitudinal relationship between Uber entry and Taxi efficiency, along with its horizontal and vertical spillover effects. To identify the entry effects on shift of transportation demands to other transportation, we rely on a natural experiment setup inherent in city Shanghai in China, as the policy is applied at same time everyday. Our empirical analyses reveal that the demand on taxi is reduced whereas demand for other alternative transportation increased. 4 - Competition in Mobile Operating Systems: the Strategy of Facebook Tongxin Zhou, University of Washington, 3927 Adams Lane NE, Seattle, WA, 98105, United States, txzhou5@uw.edu, Xue Tan React Native is a JavaScript-based cross-platform mobile development language that has attracted a lot of attention among mobile development teams. Much of the code used to develop an iOS app in React Native will work on Android and vice versa. It reduces the dependency on this very limited pool of iOS and Android platform developers and repurposes the large community of web JavaScript developers to mobile development.

213