2015 Informs Annual Meeting

SA42

INFORMS Philadelphia – 2015

SA42 42-Room 102B, CC Joint Session HAS/MSOM-Health-Practice: Operations Management of Emergency Services I Sponsor: Health Applications/MSOM Sponsored Session Chair: Maria Mayorga, Associate Professor, University of North Carolina, Dept. of Industrial & Systems Engineerin, Campus Box 7906, Raleigh, NC, 27695-7906, United States of America, memayorg@ncsu.edu 1 - Two-stage Stochastic Programming to Redeploy and Dispatch Ambulances with Restricted Workload Shakiba Enayati, Research Assistant- PhD Candidate, North carolina state university, NC State University 373 Daniels Hall, Raleigh, NC, 27695-79, senayat@ncsu.edu, Osman Ozaltin, Maria Mayorga EMS system is responsible to dispatch ambulances to arriving emergency calls. Redeployment strategy potentially improves the EMS performance as ambulances becoming busy erratically. This study proposes a stochastic approach comprising two steps. Each step is a two-stage stochastic programming in which the redeployment occurs only for idle ambulances in the first stage. Dispatching decisions are made in the second stage. Numerical results are provided based on simulation for a large real dataset. 2 - Assessing the Impact of Flexible use of Observations Units Gabriel Zayas-Caban, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States of America, gzayasca@umich.edu We assess tradeoffs that result from flexible use of Observation Units (OUs). A potential solution to reduce delays to care is to relocate boarding patients to an OU, which offer an alternative to discharging or admitting ED patients by allowing doctors to observe patients for an extended time. This allows ED beds to become available, resulting in decreased times to first treatment for patients. However, this has the potential to block the OU for patients needing observation. 3 - Dynamic Ambulance Management:Theory and Practice Rob Van Der Mei, CWI, Kruislaan 123, Netherlands, R.D.van.der.Mei@cwi.nl, Thije Van Barneveld, Sandjai Bhulai, Martin van Buuren, Caroline Jagtenberg Dynamic Ambulance Management (DAM) is a powerful means to reduce response times for ambulance services, and the use of DAM is rapidly gaining momentum. Over the past few years, we have developed a variety of DAM- algorithms, each with their pros and cons. Recently, we have started a real-life pilot to evaluate the different algorithms in practice. In this talk I will give an overview of the algorithms developed, and discuss the lessons learned from the DAM-pilot. 4 - Modeling Ambulance Dispatch Systems During Extreme Weather Events Eric Dubois, PhD Student, University of Wisconsin-Madison, 1513 University Avenue, Madison, WI, 53706, United States of America, edubois2@wisc.edu, Laura Mclay Ambulance dispatch models traditionally focus on steady state systems operating under normal conditions. We develop a Markov decision process to model the system during extreme weather events where patient queueing and patient health deterioration is relevant. We determine that under certain situations with high ambulance utilization, average patient survival can be increased by withholding ambulances from less serious patients in the expectation of more emergent future calls. Revenue Management and Learning I Sponsor: Revenue Management and Pricing Sponsored Session Chair: He Wang, MIT, 77 Mass Ave, E40-149, Cambridge, MA, 02139, United States of America, wanghe@mit.edu 1 - Learning via External Sales Networks Ankur Mani, NYU, 44 W 4th St, New York, NY, United States of America, amani@stern.nyu.edu, Josh Reed, Ilan Lobel We consider the problem of demand learning faced by a firm selling through an external sales network. The firm is not able to control its product experimentation and needs to rely on the decisions made by its sales agents. The only control available to the firm is to remove products from its lineup. We show that if the firm utilizes a well designed policy, it is able to obtain near-optimal experimentation when the sales force is sufficiently large. SA43 43-Room 103A, CC

2 - Learning and Pricing using Thompson Sampling He Wang, MIT, 77 Mass Ave, E40-149, Cambridge, MA, 02139, United States of America, wanghe@mit.edu, Kris Johnson Ferreira, David Simchi-levi We consider a network revenue management problem where a retailer aims to maximize revenue from multiple products with limited inventory. As common in practice, the retailer does not know the expected demand at each price and must learn the demand information from sales data. We propose an efficient and effective dynamic pricing algorithm, which builds upon the Thompson sampling algorithm used for multi-armed bandit problems by incorporating inventory constraints into the pricing decisions. 3 - Nonparametric Self-adjusting Price Control Stefanus Jasin, Stephen M. Ross School of Business, University of Michigan, Ann Arbor, MI, United States of America, sjasin@umich.edu, George Chen, Izak Duenyas We consider dynamic pricing of multiple products with limited inventories. The functional form of demand is not known. We devise a nonparametric heuristic that consists of four elements: Spline approximation of the unknown demand during the exploration stage, linear approximation of the estimated demand, quadratic approximation of the estimated revenue, and self-adjusting control during the exploitation stage. Our heuristic significantly improves the theoretical bound of existing heuristics. 4 - Incomplete Learning and Certainty-equivalence Control Bora Keskin, Duke University, Fuqua School of Business, 100 Fuqua Drive, Durham, NC, 27708-0120, United States of America, bora.keskin@duke.edu, Assaf Zeevi Motivated by dynamic pricing applications, we consider a dynamic control and estimation problem where a system manager sequentially chooses controls and makes observations on a response variable that depends on chosen controls and an unknown sensitivity parameter. The system manager uses a certainty- equivalence decision rule to determine subsequent controls based on estimates, and we characterize the asymptotic accuracy performance of this policy. Pricing and Consumer Behavior Sponsor: Revenue Management and Pricing Sponsored Session Chair: Mikhail Nediak, Queen’s University, 143 Union Str., Kingston, ON, K7L3N6, Canada, mnediak@business.queensu.ca Co-Chair: Yuri Levin, Queen’s School of Business, 143 Union St. West, Kingston, Canada, ylevin@business.queensu.ca 1 - A Non-parametric Approach to Dynamic Pricing with Demand Learning Guyves Achtari, Queen’s School of Business, 143 Union Str., Kingston, ON, Canada, 11ga10@queensu.ca, Mikhail Nediak In many industries, firms have the capability of observing both sales and the refusal to buy from their customers. In situations where demand is unknown, firms may use early sales data to forecast demand. We consider a situation where the firm does not know demand, but can observe arriving customers refuse or accept to buy a product at a given price. We formulate a dynamic program which aims to dynamically adjust the price of the product in order to maximize the firm’s total expected revenue. 2 - Turnpike Equilibrium for Oligopolistic Dynamic Pricing Competition with Strategic Consumers Jue Wang, Post-doctoral Fellow, Queen’s School of Business, 143 Union St. West, Kingston, ON, K7L 3N6, Canada, jw171@queensu.ca, Yuri Levin, Mikhail Nediak We consider the oligopolistic price competition when the prices are dynamic and the customers are strategic. We formulate the problem as a fluid model in the optimal control framework, and show that the equilibrium has a turnpike property. We characterize the structure of turnpike for symmetric oligopoly and asymmetric duopoly. The impact of non-stationary demand is also examined. 3 - Quantity Competition in a Multi-product Exchange Market with Strategic Consumers and Dynamic Preference Samuel Kirshner, UNSW Business School, Kensington, NSW, Australia, skirshner@business.queensu.ca, Mikhail Nediak, Yuri Levin We study a general multi-product multi-period exchange market for gross substitute products. Consumers maximize surplus under the uncertainty of their future product preferences deciding the quantity of products to purchase and sell in each period. Under mild assumptions, the equilibrium trades and price path is unique. The model is used to explore how strategic behavior of consumers and preference dynamics impact the equilibrium and aggregate welfare. SA44 44-Room 103B, CC

52