Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TD39

3 - Joint Pricing and Matching in Ridesharing Systems Erhun Ozkan, Koç University, Istanbul, Turkey

consider two models, corresponding to the two dependence structures, to which we refer as exogenous and endogenous dependence. To approximate the intractable queueing processes, we propose a unified fluid approximation which captures two models simultaneously while exposes fundamental differences between the two models.

Two key questions common to all ridesharing firms concern pricing and matching. The pricing decisions determine both customer demand and driver supply. The decision of which driver to match to an arriving customer determines how long that customer must wait for driver pick-up. We study the interplay between the pricing and the matching decisions of a ridesharing firm. We show that optimizing only in the pricing (matching) dimension while using same area matchings (constant prices) has no benefit to the firm. Therefore, pricing and matching decisions should be optimized jointly. 4 - On Matching and Thickness in Heterogeneous Dynamic Markets We study dynamic matching in an infinite-horizon stochastic market. While all agents are potentially compatible with each other, some are hard-to-match and others are easy-to-match. Agents prefer to be matched as soon as possible and matches are formed either bilaterally or indirectly through chains. We adopt an asymptotic approach and compute tight bounds on the limit of waiting time of agents under myopic policies that differ in matching technology and prioritization. n TD39 North Bldg 226A Queueing and Inventory Models Sponsored: Applied Probability Sponsored Session Chair: John Hasenbein, University of Texas-Austin, Austin, TX, 78712- 0292, United States 1 - Stochastic Analysis of Queues with Information Updates Jamol Pender, Cornell University, 228 Rhodes Hall, Ithaca, NY, 14850, United States Many service systems provide real-time information to their customers with the goal of reducing the customers’ anxiety of the unknown. However, the information might be unreliable and is often not given in real-time. In this talk, we will show how to prove fluid limit theorems for a state dependent infinite server queueing model where customers choose which queue to join by a generalized customer choice model. In the choice model, the information about the queue length is updated in discrete intervals unlike a constant delay model. We show that the fluid limit is given by a system of functional differential equations with a non-stationary time delay. 2 - Queue Length Asymptotics for the G/G/D Queue with Heavy-tailed Service Times Chang-Han Rhee, Centrum Wiskunde and Informatica, Jakoba Mulderplein 164, Amsterdam, 1018 MZ, Netherlands, Mihail Bazhba, Jose Blanchet, Bert Zwart In this talk, we answer the long-standing open question posed by Sergey Foss and Ward Whitt regarding the queue length asymptotics of the multiple-server queues with heavy-tailed service times. It is well documented that queueing systems typically experience congestions because of a small number of exceptionally large jobs in heavy-tailed environments. However, the rigorous characterization of such phenomenon has been limited to power-law type service times. In this talk, we show that when the service time distribution has a Weibull tail, a number of unexpected and interesting features—-such as the asymmetry in job sizes and the trade-off between the number of big jobs and their sizes—-arise. 3 - Optimal Stock Allocation for Production-inventory Systems with Multiple Impatient Customer Classes Yasar Levent Kocaga, Sy Syms School of Business, Belfer Hall Room # 403/A, 2495 Amsterdam Ave, New York, NY, 10033, United States, Yen-Ming Lee We address the production and inventory control of a make-to-stock system with multiple impatient customer classes. We assume Poisson demand and exponential production times. Demand not satisfied immediately is backordered; but waits only up to an exponentially distributed amount of time and is cancelled if not satisfied within this time. We show that the threshold inventory rationing policy, which is optimal under backordering and lost sales, is still optimal under certain conditions including a requirement on the order of order cancellation rates. 4 - A Unified Fluid Model for Service Systems with Exogenous and Endogenous Dependencies Chenguang (Allen) Wu, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong, Achal Bassamboo, Ohad Perry A prevalent assumption in the queueing literature is that the service times of customers are independent of the system’s state, and of all other primitives comprising the system. However, in many practical settings, the service requirements of arriving customers may depend on their individual patience times, or on the delay they experienced while waiting to be served. We thus Vahideh Manshadi, Yale University, New Haven, CT, 06511, United States, Itai Ashlagi, Maximilien Burq, Patrick Jaillet

n TD40 North Bldg 226B

Spatial Decision Analysis Sponsored: Decision Analysis Sponsored Session Chair: Valentina Ferretti 1 - Persistent Patterns of Discriminatory Housing Policy and Inequity: A Spatial Statistical Analysis Anna White, University of Michigan, Ann Arbor, MI, United States, Seth Guikema Racial discrimination by housing and lending institutions was legal and widely practiced well into the 1960s in the United States. Although such discrimination is now illegal, the footprint of the system can be visualized through discriminatory “redlining maps created by the Home Owner’s Loan Corporation (HOLC). We use Bayesian Belief Networks and statistical methods to analyze the relationships between maps created by the HOLC and current day quality-of-life indicators like health, income, employment, and education in cities across the United States. Through this analysis, we reveal the patterns that endure across the country as a result of this system. 2 - Weight Elicitation for Spatial Decision Problems Jay Simon, American University, 4400 Massachusetts Avenue, NW, Washington, DC, 20016, United States Decisions with spatial outcomes present several preference assessment challenges. Some can be addressed via preference conditions that make the functional forms of value and utility functions simpler. Others, however, are unavoidable; if an outcome will occur across a large geographic space, then it is necessary to elicit a large number of weights from stakeholders. We present a fast and cognitively simple weight approximation technique for a large number of regions, and show Valentina Ferretti, London School of Economics and Political Science, Houghton Street| London | WC2A 2AE, London, United Kingdom The need and interest to consider cognitive and motivational biases has been recognized in different disciplines (e.g. economics, decision theory, finance, risk analysis) and has recently reached environmental decision making. Within this domain, Spatial Decision Analysis represents a particularly interesting domain to explore new dimensions and implications of behavioral aspects. This talk will present insights from a first behavioral experiment on the spatial dimension of biases. The aim of the experiment is to understand whether or not and to which extent the use of spatial information (maps) can bias the weights’ elicitation phase of a Multicriteria Decision Aiding process. n TD41 North Bldg 226C Cyber Risk and Public Policy Sponsored: Decision Analysis Sponsored Session Chair: Jonathan W Welburn, RAND Corporation, Pittsburgh, PA, 15213, United States 1 - Text Mining and Public Policy Seifu John Chonde, Penn State University, Alexandria, VA, 22314-2299, United States RAND is devoted to helping the public and clients understand the impact of policies. Over the past few years, RAND has developed a suite of text analysis tools known as RAND-Lex that have been applied to a variety of different types of texts. This talk will overview a subset of research projects that RAND has completed to demonstrate the ongoing issues faced in text mining. For example, recent work has shown the utility of embedding models in natural language tasks such as sentiment analysis and document classification; however, the bias of these models is largely unknown and is an ongoing area of research complicating the use of these models for certain applications. that it performs well relative to commonly used approaches. 3 - Cognitive Biases in Spatial Decision Analysis

360