Informs Annual Meeting 2017

TE17

INFORMS Houston – 2017

5 - Decision Conflict in the Newsvendor Game Ilkka Leppanen, Dr, Loughborough University School of Business and Economics, Epinal Way, Loughborough, LE11 4NZ, United Kingdom, i.leppanen@lboro.ac.uk We argue that decision conflict between the pecuniary motive of profit maximization and the nonpecuniary motive of satisfying demand is the driving force behind non-normative behavior in the newsvendor game. We find in behavioral experiments that the previous round situation affects current round decision time such that “intermediate” situations where demand was not satisfied and the decision was not normative produce more decision conflict and deliberation in the current round than “extreme” situations where either demand was satisfied or/and the decision was normative. We demonstrate that decision conflict between pecuniary and nonpecuniary motives can affect newsvendor behavior. 6 - Is the Optimal Suggestion the Best for Managerial Decision Making? A Laboratory Investigation of the Newsvendor Problem Xiaojing Feng, PhD Candidate, Shanghai Jiao Tong University, 2nd Yanjiusheng Building 1519, Huashan Road 1954, Shanghai, 200031, China, fengxj89@sjtu.edu.cn, Ying Rong, Tianjun Feng Decision makers often tend to choose the inferior humans’ forecasts rather than the superior algorithms’ forecasts, although algorithms outperform humans in most cases. So, how to design an efficient decision support system (DSS) that helps people to make better forecasts and decisions becomes our concern. We investigate how the suggestions of different DSSs affect humans’ order decisions in the classical newsvendor settings. Through laboratory experiments, we find that the introduction of DSS induces human newsvendors to exhibit algorithm aversion and regret aversion. We identify the most efficient DSS which eliminates human newsvendors’ pull-to-center bias. 340A Service Systems in Applied Probability Sponsored: Applied Probability Sponsored Session Chair: Jing Dong, Northwestern University, Evanston, IL, 60208, United States, jing.dong@northwestern.edu 1 - A Fluid Model for Service Systems with Dependent Service and Waiting Times Chenguang Allen Wu, Northwestern University, Evanston, IL, United States, allenwu@u.northwestern.edu, Ohad Perry, Achal Bassamboo We consider a large service system in which the service time of each customer depends on his waiting time. Specifically, there can be two types of dependence: either a customer’s patience depends on his individual service requirement, or the actual service requirement changes in response to the delay in queue. To study the two cases, we develop a unified fluid model that captures those two types of dependencies, and demonstrate its accuracy via simulations. We prove qualitative results for the stationary and transient behavior of the fluid model, which reveal the fundamental differences between the two dependence structures. 2 - Multi Queue Service Systems with Dynamic Customer Choice Yichuan Ding, University of British Columbia, University of British Columbia, 6333 Larkin Drive, Vancouver, BC, V6T.1C3, Canada, daniel.ding@sauder.ubc.ca, Mahesh Nagarajan, Zhe Zhang We study a stochastic service system with multiple servers and dynamic consumer choice., where service providers of heterogeneous quality are distributed at different locations. An arriving customer chooses a service provider to obtain service by observing the service reward and length of each queue. Many real life queueing systems can be accommodated by this model. We study this system using both fluid and diffusion approximations. We prove the existence of the unique equilibrium customer queue-length vector of the service providers.We show that the system can be approximated by a multi-dimensional reflected TE17

4 - Optimal Policy in Single-server Multi-class Queuing Systems with Abandonment Sina Ansari, Northwestern University, Evanston, IL, United States, SinaAnsari2013@u.northwestern.edu, Seyed Iravani, Laurens G. Debo We study a multi-class queuing system with a single server and customer abandonment. We fully characterize the structure of the server’s optimal scheduling policy that minimizes the total average customer abandonment cost. 340B Queueing Systems and Approximations Sponsored: Applied Probability Sponsored Session Chair: John Hasenbein, University of Texas-Austin, Austin, TX, 78712-0292, United States, jhas@mail.utexas.edu 1 - An Asymptotically Optimal Control Policy for a Multiclass Many Server Queue with General Reneging Distributions Amy R. Ward, University of Southern California, Marshall School of Business, Bridge Hall BRI 401H, Los Angeles, CA, 90089-0809, United States, amyward@marshall.usc.edu, Amber L. Puha We consider a multiclass many-server queue with general interarrival, service, and reneging distributions. For reneging distributions with bounded, nonincreasing hazard rates, we find that static priority may not be asymptotically optimal on fluid scale. We propose a new class of policies, called Random Buffer Selection, and prove their asymptotic optimality. We further identify the fluid approximation to the limiting cost as the optimal value of a certain optimization problem. 2 - Operational Benefits of Free Trials in Large Scale Service Systems Yasar Levent Kocaga, Sy Syms School of Business, Belfer Hall Room # 403/A, 2495 Amsterdam Ave, New York, NY, 10033, United States, kocaga@yu.edu, Chihoon Lee We consider a firm that is serving price and delay sensitive customers and that has the option of offering a free trial service to a new market of customers. We first show that it is optimal for the revenue maximizing firm to operate in the QED heavy traffic regime and provide tractable and accurate expressions for the optimal price and revenue. Then, we use these expressions to identify conditions under which offering free trials is beneficial to the firm. 3 - Stability and Tail Asymptotics in a Multiclass Queue with State-dependent Arrival Rates John Hasenbein, University of Texas-Austin, 1 University Station Stop C2200, Department of Mechanical Engineering, Austin, TX, 78712-0292, United States, jhas@mail.utexas.edu, Philip Ernst, Soren Asmussen We analyze a multiclass single-server queueing system in which the arrival rates depend on the current job in service. The system is characterized by a matrix of arrival rates. Using fluid models, we obtain the necessary and sufficient conditions for stability. Utilizing the natural connection with the multitype Galton-Watson processes, we derive the Laplace-Stieltjes transform of busy periods. We conclude with tail asymptotics for the busy period for heavy-tailed service time distributions for the regularly varying case. 4 - Naor’s Model with Heterogeneous Customers and Arrival Rate Uncertainty Chengcheng Liu, 2815 Rio Grande Street, Austin, TX, 78705, United States, liuchengcheng2011@gmail.com, John Hasenbein This research examines extensions of Naor’s queueing model in which the arrival rate is not known with certainty by either customers or system managers. The arrival rate is considered as a non-degenerate random variable. In view of a social optimizer and a revenue maximizer, we investigate both static and state- dependent pricing policies in the presence of customer heterogeneity in their economic characteristics. Two cases are considered: the observable case and the unobservable case. This work is also an extension of the study of Chen and Hasenbein on a related model in that their model only considers homoge- neous customer populations. TE18

Ornstein-Uhlenbeck process centered at that equilibrium. 3 - Central Limit Theorems for Bike-sharing Networks Jamol Pender, Cornell University, jjp274@cornell.edu

Bike-sharing is an emerging mode of eco friendly transportation in many cities. In this talk, we prove central limit theorems for a bike-sharing stochastic model. These central limit theorems allow us to move beyond mean field behavior and describe the fluctuations around the mean field limit. Moreover, we are able to derive new Gaussian approximations for various performance measures such as the probability that a station has no bikes or is at capacity.

416