INFORMS 2021 Program Book

INFORMS Anaheim 2021

MC45

to show how these optimization methods can be effectively combined and applied to a realistic instance. We use a recent spanning tree algorithm (Recombination) to create congressional and state legislative district plans for Missouri that are optimized for various fairness metrics. To satisfy Missouri’s legal requirements, we use edge weights and a combination of counties, census tracts, and census block groups as graph nodes. 3 - Inmate Overcrowding and Prison Operations: A Review Ben Lewis, Research Fellow, University of Michigan, Ann Arbor, MI, United States Overcrowding has been an ongoing issue in federal and state prisons for decades, and many US prisons are still operating over capacity despite efforts to reduce prison populations in the wake of COVID-19. This systematic literature review aims to 1) highlight the individual, group, and societal impacts of inmate overcrowding, 2) illustrate the current landscape of prison operations literature, and 3) identify optimization approaches that could reduce overcrowding in prisons. Co-citation analysis suggests that overcrowding generally has negative impacts on prisoners and prison operations and that assignment and queuing approaches are best suited for minimizing prison populations. 4 - Effective, Fair and Equitable Pandemic Rationing Aram Grigoryan, Duke University, Durham, NC, United States We study the problem of allocating scarce and heterogeneous medical resources such as COVID-19 vaccines. Our solutions optimize for aggregate match- effectiveness subject to fairness and distributional constraints. The main solution uses a novel cutoff characterization result for fair and equitable allocations and a minimum-cost flow formulation. Match-effectiveness gains from optimization are substantial. Even when there are only two types of vaccines, in equal quantities, our algorithm results in more than 33% larger aggregate match-effectiveness compared to the random allocation benchmark. 5 - The Cost of “Kung Flu”: Negative Social and Economic Impacts of Anti-China Rhetoric on Asian Americans Justin T. Huang, Assistant Professor of Marketing, University of Michigan Ross School of Business, Ann Arbor, MI, United States, David Rothschild, Julia Lee Cunningham, Francesca Gino, Masha Krupenkin The past year saw a rise in hate directed towards Asian Americans in the US. We track these events back to the onset of Covid-19 and statements directing blame to China (‘Kung Flu’). In a series of analyses combining surveys, online search trends, and consumer location data, we show how biases intensified and led to changes in both search and choice behavior. Survey respondents rated Asians as less American than other groups, and searches for stereotypes around Asian restaurants spiked. Asian restaurant traffic dropped 10.9% relative to non-Asian restaurants. We explore heterogeneity in these effects and identify ethnic misidentification as a driver of spillovers to non-Chinese Asian restaurants. MC43 CC Room 213A In Person: Computing/Auctions/Mechanism Design Contributed Session Chair: Manxi Wu, University of California, Berkeley, CA, United States 1 - Learning-based Resource Management for Mobile Edge Computing Systems Hana Khamfroush, University of Kentucky, Lexington, KY, United States, Sam Heshmati With the growing needs of real-time data analytics, mobile edge computing (MEC) is becoming a popular technology to process large scale data at the edge of the system and close to the users. MEC however comes with its own limitations such as limited computing, communication, and storage resources. Therefore, smart resource management strategies are needed to provide efficient use of these resources. This talk will address the use of deep learning models for resource management in the mobile edge computing systems. Challenges and opportunities are presented, while discussing some preliminary results. 2 - Numerical Solutions to a Fredholm Form Of Integral Equations for Finite Measures Shukai Li, Northwestern University, Evanston, IL, United States, Sanjay Mehrotra We study a form of integral equations for finite measures, which arise in many applications including stationary distribution problems and Markov chain Monte Carlo. We exploit the properties of our equations and apply a discretization approach for approximate solutions. Specifically, we construct a Banach space of distribution functions to reformulate the problem into a Fredholm-form operator equation and outline necessary and sufficient conditions for applying collective compactness theory. We provide convergence results for the discretization approach and analyze how to compute the approximate solutions as well as their error bounds via a linear program under appropriate assumptions. 3 - Scarcity and Waste in Allocation Mechanisms Junxiong Yin, University of Southern California, Los Angeles, CA, United States, Peng Shi

Variants of wait-lists are used to allocate scarce resources such as cadaver kidneys. However, around 20% of successfully procured cadaver kidneys are discarded. In this paper, we study the wastage problem from a theoretical perspective with a focus on the wait-list with choice, which is an approximation to the current mechanism for cadaver kidney allocation. We find that 1) it is not always possible to Pareto improve upon the wait-list with choice even when there is waste; and 2) it is impossible to Pareto improve upon the wait-list with choice using a mistake-tolerant mechanism. The findings suggest that reducing waste requires hard discussions among stakeholders. MC44 CC Room 213B In Person: Supply Chain Optimization Contributed Session Chair: Elham Taghizadeh, Wayne State University, Clinton Township, MI, 48035-5630, United States 1 - Multi-Year, Multi-Commodity Supply Chain Network Design Seyed Mohammad Nourbakhsh, Walmart, San Bruno, CA, United States, Seth Kim, Willie Montgomery We developed a two-step heuristic optimization model that minimizes the total network cost for SAM’s club supply chain network. The proposed model determines the optimum plan of future expansion, new fulfillment construction, and other decisions for the next 10 years. 2 - Responsive Production Planning and Replenishment Scheduling for a Two-echelon Supply Chain Sepideh Alavi, California State University-San Bernardino, San Bernardino, CA, United States In this research, we study the integrated production and inventory replenishment problem for a two-echelon supply chain. The literature on replenishment planning focuses on specific types of replenishment policies. Such policies have operational shortcomings in a sense that they do not reflect the flexibility of adjusting the replenishment schedules based on changing market conditions. We will consider the problem of obtaining a detailed responsive replenishment plan over a planning horizon. 3 - A Dynamic Resilience Management Framework for Deep-tier Automotive Supply Networks Elham Taghizadeh, Wayne State University, Clinton Township, MI, United States, Ratna Babu Chinnam, Saravanan Venkatachalam We propose a framework to manage the resilience of deep-tier automotive supply networks. We integrate a simulation-based resilience assessment scheme with an efficient optimization-based framework for resilience management. The framework promotes the use of network analysis techniques combined with discrete-event simulation informed by secondary data sources and global supply risk databases for improving resilience management. We validate the effectiveness of the proposed framework using a global automotive OEM case study. MC45 CC Room 213C In Person: Production & Scheduling Contributed Session Chair: Stanislaus Solomon, Sam Houston State University, Huntsville, TX, 77341, United States 1 - An Agent-based Approach to the Job Shop Scheduling Problem with Order Rejection Omar Abbaas, Graduate Assistant, Pennsylvania State University, University Park, PA, United States, Jose Antonio Ventura, Sara Abu Aridah, Kevin Bunn This study considers the job shop scheduling problem with order rejection and earliness and tardiness penalties using an agent-based approach with a combinatorial auction mechanism. A set of jobs is offered. Each job has a revenue, ready time, due date, deadline, and consists of a set of operations with unique precedence relationships. A mathematical model is presented, then Lagrangian relaxation is used to decompose the problem into a set of job-level scheduling problems. Profitable jobs at the individual level submit their bids to an auctioneer. Then, the auctioneer resolves conflicts to reach a feasible solution, records the profit upper and lower bounds, and updates the dual variables.

71