INFORMS 2021 Program Book

INFORMS Anaheim 2021

MB43

5 - Impact and Risk Models on COPD-related Hospitalizations and Emergency Room Visits by Combining Multi-year Claims Data with Environmental Data Sets Divya Mehrish, CapsicoHealth Intern; Stanford University Student, CapsicoHealth, Palo Alto, CA, United States, J. Sairamesh, Laurent Hasson, Monica Sharma, Rudy Banerjee, Jakob Bjorner Chronic obstructive pulmonary disease (COPD) is the fourth-leading cause of death in the U.S. Our study examines clinical and environmental impacts on 90- day hospitalizations, ER visits and readmissions. We integrate 2017-18 CMS claims data with daily weather and pollution data in FL, NY and PA (high annual hospitalization rates). Our logistic regression models show 88% accuracy for 90- day hospitalizations and 74% for ER visits; 94% and 79% with boosted tree-based models. Our models all show 64% accuracy for 30-day readmissions. Our results, which show the clear relationship between the environment and COPD hospital and ER cases, can help care managers target high-risk populations. MB43 CC Room 213A In Person: Computer Science Applications to OR Contributed Session Chair: Mesut Yavuz, University of Alabama, Tuscaloosa, AL, 35487, United States 1 - Exactly Solving Linear Systems via the Sparse Exact (SPEX) Framework: History and Theoretical Foundation Erick Moreno-Centeno, Texas A&M University, College Station, TX, United States, Christopher Lourenco Solving sparse linear systems has a central role in solving linear programs and other optimization problems. Exactly solving linear programs and systems is necessary for some applications (e.g., theoretical results, feasibility problems, military applications, applications with hefty costs, ill-conditioned problems, etc.). To address this, we are developing the Sparse Exact (SPEX) Factorization Framework: a high-performance, well-documented, and extremely robust software package. This talk will focus on the history and the theoretical foundations of the package, and a companion talk by Christopher Lourenco will focus on the recent developments and computational results. 2 - Exactly Solving Linear Systems via the Sparse Exact (SPEX) Framework: Moving Towards Exact Optimization Christopher Lourenco, Assistant Professor, U S. Naval Academy, Annapolis, MD, United States, Erick Moreno-Centeno Solving sparse linear systems, via LU, Cholesky, and other factorizations, is a fundamental subroutine in mathematical programming. Though most solvers operate exclusively in double precision; applications where more precision is needed are increasingly forcing solvers to move towards quad precision or even fully exact solutions. This talk presents a framework to exactly solve sparse linear systems like those in mathematical programming. Our presented algorithms operate exclusively in integer-arithmetic and we provide computational results showing that they outperform the alternate exact approaches of rational- arithmetic and exact iterative methods. 3 - Throughput-fairness Tradeoffs in Mobility Platforms Arjun Balasingam, Massachusetts Institute of Technology, Cambridge, MA, United States, Karthik Gopalakrishnan, Radhika Mittal, Venkat Arun, Ahmed Saeed, Mohammad Alizadeh, Hamsa Balakrishnan, Hari Balakrishnan We study the problem of scheduling, routing, and allocating tasks from different customers to vehicles in shared mobility platforms (e.g., food and package delivery, ridesharing, and mobile sensing). We introduce Mobius, a system that uses guided optimization to navigate the inherent tradeoffs between fairness and throughput caused by shared mobility. Mobius supports spatiotemporally diverse and dynamic customer demands. Our evaluation demonstrates these properties, along with the versatility and scalability of Mobius, using traces gathered from ridesharing and aerial sensing applications. 4 - A Study of Software Development Practice in Operations Research Mesut Yavuz, University of Alabama, Tuscaloosa, AL, United States, Huseyin Ergin Software is a crucial part of operations research (OR). In this talk, we present the results of two studies. The first explores all papers published in INFORMS Journal on Computing in the 5-year window (2016-2020) and the second one is a survey of 389 OR scholars. The results shed light on the current state of software development practice in OR, reveal the relationship of the OR scholars with research software, and present the expectations and concerns of them regarding code and data sharing practices.

MB44 CC Room 213B In Person: Supply Chain Management III Contributed Session Chair: Yasamin Salmani, Bryant University, Smithfield, RI, 02917-1220, United States 1 - Data-Driven Distributionally Robust Supply Chain Contracting with Stock-Out Substitution Xuejun Zhao, Purdue University, West Lafayette, IN, United States, William Haskell We study supply chain contracting with stock-out substitution, when the supplier only has partial information about demand distribution obtained from the past demand realizations and retailer’s ordering decisions. We propose a distributionally robust contract for the supplier to hedge against the risks of extremal demand distributions. The uncertainty set combines the information from the retailer’s ordering decisions and the past demand realizations, based on the Wasserstein distance. We will show both analytical and computational properties of our new uncertainty set. 2 - Weather Rebate Contracts with Buyback Policy Piyal Sarkar, PhD Candidate, Ryerson University, Toronto, ON, Canada, Mohamed Wahab Mohamed Ismail, Liping Fang Firms dealing with weather sensitive products often face a problem with demand management. A class of contract for a supplier-retailer supply chain to address this issue is proposed. This contract provides an incentive to the retailer in the form of a weather rebate to induce the retailer to increase the ordering quantity and takes care of the inventory risk by an inventory buyback policy. The supplier uses weather derivatives to hedge risk depending on the risk attitude. CVaR is used to model the risk attitude. The results show that the designed contract performs better than a traditional buyback contract. The study designs a new class of contract that can be used to manage a supply chain under weather risk. 3 - Supply Chain Viability During Pandemic Yasamin Salmani, Bryant University, Smithfield, RI, United States, Amin Ariannezhad Companies respond differently to the pandemic: some sink into bankruptcy, some resist the shock, and some get better. In this study, utilizing a data-driven approach, we investigate the impacts of various established supply chain partner models on the resilience and viability of the companies in the context of the COVID-19 pandemic. We use the companies’ performance data before and during the pandemic to address this problem. MB45 CC Room 213C In Person: Manufacturing and Inventory Management Contributed Session Chair: Erik Bertelli, University of California-Berkeley, Alameda, CA, 94501, United States 1 - Resource Allocation of Inspections in Genetic Manufacturing Systems by Using MDP Approach Mohammad Maydanchi, Auburn University, Auburn, AL, United States, Gregory T. Purdy, Daniel F. Silva Genetic Manufacturing System (GMS) is a new type of manufacturing with a genetic construct as the final product. In GMS, like other manufacturing systems, having a defective outcome increases the cost and time of the operation. Mid- process inspection is used to check the quality, but the type and number of inspections could adversely affect the time and cost of the product. This work deploys a Markovian Decision Process (MDP) approach to indicate preferred inspection strategies to minimize the total cost and improve the quality based on the non-conforming rate of operations and Type I and Type II error rates. 2 - Manufacturing Localization and its Performance Implications: An Empirical Study in the Automotive Industry Zhenzhen Yan, Ph.D. Candidate, Michigan State University, East Lansing, MI, United States, Sriram Narayanan, Tobias Schoenherr, Sourish Sarkar Literature on manufacturing relocation focuses on the decision drivers while the decision consequences are understudied due to the difficulty of data collection. This study contributes to this stream of literature by investigating the performance implications of manufacturing localization, firms’ decision of relocating manufacturing activities closer to the target market. Specifically, we use the automotive industry as an empirical context and apply a causal estimation to a uniquely assembled dataset across recalls, inventory, and other industry-specific proprietary data compiled over a 20-year period. Our findings provide practical insights to firms that consider relocation.

56