INFORMS 2021 Program Book

INFORMS Anaheim 2021

WE21

3 - Optimizing Vehicle Fleet and Assignment for Concentrating Solar Power Plant Mirror Washing Alexander Zolan, National Renewable Energy Laboratory, Austin,

2 - Appointment Scheduling with Multiple Servers and Optional Batch Arrivals Robert H. Lee, University of Amsterdam, Amsterdam, Netherlands, Alex Kuiper In healthcare appointment scheduling one often seeks to minimize a weighted sum of patient waiting time and physician idle time. It is common to suppose that patients arrive singly and that the schedule is served by a single physician. However, these assumptions are frequently relaxed. By formulating an appointment schedule as a queue where patients arrive in batches of size at least one, and where there are one or more servers, a Db/M/c queue, one can analytically investigate the effect on optimal appointment schedules of varying the batch size and the number of servers independently of one another. 3 - VRP with Release Dates and Deadlines: A Blood Sample Collection Application Fernando A.C. Fontes, Universidade do Porto, Faculdade de Engenharia, Universidade do Porto, NI, Porto, 4200-465 PORTO, Portugal, Dalila B. M. M. Fontes, Helena V. Ferreira Blood sample collection is critical due to the blood short lifespan. Once extracted, the blood samples are stored until collected, transported, and delivered to the lab. Biological degradation imposes a limit on the time between extraction and delivery. If this limit is not respected, the blood must be disposed of and a new extraction arrange for, which in addition to the extraction, collection, and delivery costs, also implies environmental costs associated with the disposal of biological residues. This problem can be cast as a VRP with release dates (the extraction timing) and deadlines (the lifespan). We propose a MILP model to solve this problem. WE21 CC Room 204A In Person: Marketplace Design and Operations General Session Chair: Park Sinchaisri, The Wharton School, University of Pennsylvania, Philadelphis, PA, United States 1 - Crowdsourcing Market Information From Competitors Irene Y Lo, Stanford University, Stanford, CA, 94305-4121, United States, Joann de Zegher Market price information is often not widely available in the developing world, and information sharing agreements among competing firms can create significant benefits. However, such agreements may be difficult to implement, as a firm might fear that sharing information will benefit competitors. We show that an appropriately designed information-sharing platform can disclose partial information that will benefit all firms. By eliminating business stealing concerns, our information disclosure policy creates a Pareto improvement and is implementable if the information shared by the platform is sufficiently valuable. The model requires minimal assumptions and can account for general market dynamics. The interpretability of our results allows us to propose a heuristic for use in practice by an Indonesia-based information-sharing platform we collaborate with. 2 - Intertemporal Pricing with Resellers: An Empirical Study of Product Drops Dayton T. Steele, University of North Carolina Chapel Hill, Ashbrook Apts 601 Jones Ferry Rd Apt B7, Carrboro, NC, 27510- 2159, United States, Seyedmorteza Emadi, Saravanan Kesavan Product drops occur when a firm releases a limited-edition product line on a specific date for a short period of time. The product drop generates hype from customers that results in large sales, and a resale market may emerge where products resell at higher prices once the firm stocks out. A firm may ask: “Am I leaving money on the table?” To answer this, we obtain a unique data set from a retailer of baby clothing with weekly product drops. We estimate a structural model that incorporates the strategic behavior of customers reselling as well as firm pricing decisions based on limited inventory. We find that ignoring the resale market in pricing reduces firm profit by 7.0% on average. 3 - The Structural Behavioral Model of Gig Economy Workers Park Sinchaisri, Assistant Professor, Haas School of Business, UC Berkeley, Berkeley, CA, 19104, United States, Gad Allon, Maxime Cohen, Kenneth Moon With the flexibility in the choice of service, gig workers often exhibit a “multihoming” behavior. An increase in the number of options available to gig workers has resulted in increased competition among platforms to win over a limited mutual pool of workers. How workers respond to platform competition is therefore an important topic to study, but studying multihoming behavior empirically is challenging due to the unobservability of work options. We combine proprietary data from a ride-hailing platform and public trip records to estimate a structural model of workers’ labor decisions when facing multiple work opportunities, using a machine learning-based adversarial estimation approach. Our counterfactual analyses demonstrate the effectiveness of different policies and offer insights that can help the firm manage its workers during different demand scenarios.

TX, 78757-2608, United States, Jesse Wales, Alexandra M. Newman, Michael J. Wagner

Concentrating solar power central-receiver plants use thousands of sun-tracking mirrors, i.e., heliostats, to redirect sunlight to a central receiver, which collects and uses the heat to generate electricity. Over time, soiling reduces the reflectivity of the heliostats and, therefore, the efficiency of the system. We present a mixed- integer nonlinear program that determines wash vehicle fleet size, mix, and assignment of wash crews to heliostats to minimize the sum of (i) the revenues lost due to soiling, (ii) the costs of hiring crews and operating vehicles, and (iii) the costs of purchasing vehicles. We propose a decomposition method that enables near-optimal solutions to the wash vehicle sizing and assignment problem on the order of a couple of minutes. These solutions yield hundreds of thousands of dollars in savings per year over current industry practices.

WE19 CC Room 203A In Person: Studies of the Pharmaceutical

Supply Chain General Session Chair: Emily L Tucker, Clemson University, Clemson, SC, 29634, United States 1 - The Interactions of Crowding, Patient Severity, and Queue Rank at a Hospital Emergency Department Lu Wang, Ball State University, Muncie, IN, United States, Mazhar Arikan, Suman Mallik Utilizing the patient data from the ED of a large urban teaching hospital, we characterize the impacts of the change in patient queue rank on patient LOS. We study how arrivals of higher/lower severity patients influence patient LOS, and how changes in queue rank, severity, and crowding simultaneously affect LOS. 2 - International Drug Shortages: Associations Between Shortages Across Continents Drug shortages regularly occur around the globe. Beyond the immediate concerns of COVID-19 supply and demand pressures, supply chain issues have caused shortages for decades. Much of the research about drug shortages in the US has abstracted away the global context of these international supply chains. European researchers have begun to consider the effects of shortages across intra- continental borders, and in this work, we evaluate the associations between shortages in the US and other countries around the world. We consider whether particular causes increase the likelihood of occurrence in other countries and discuss potential ramifications for policy. WE20 CC Room 203B In Person: Health Care, Modeling and Optimization II Contributed Session Chair: Fernando A C C Fontes, Universidade do Porto, Porto, 4200-465, Portugal 1 - Temporal Network Architectures of Neurocognitive and Psychological Symptoms in Collegiate Athletes with Sport-related Concussion Caroline G. Turner, United States Naval Academy, Annapolis, MD, United States, Anna Svirsko, Gian-Gabriel P. Garcia, Spencer Liebel Concussions are a common brain injury, affecting millions of Americans each year but, the relationship between concussion symptoms during the healing process is not well understood. In order to further understand the concussion recovery process, we develop a weighted temporal network to analyze how concussion symptoms are interrelated, mutually reinforcing, and amplifying. In analyzing this network, we look to identify each symptom’s evolution through the healing process and how symptoms influence each other over to allow for a better understanding of the concussion recovery process. Emily L. Tucker, Assistant Professor, Clemson, SC, 29634, United States, Emilia Vann Yaroson, Shravan Anil Shinde, Martha L. Sabogal De La Pava

175