INFORMS 2021 Program Book

INFORMS Anaheim 2021

TE17

2 - Incorporating Ventilation, Heat, and Emissions in a Short-term Underground Mine Production Scheduling Model John Ayaburi, MS, Colorado School of Mines, Golden, CO, 80401, United States Mine planners utilize production schedules to determine when blocks of ore should be extracted. However, the accumulation of heat in an underground mine not only disrupts the schedule but also affects the health and safety of mine workers. We propose a large-scale, short-term production scheduling model that minimizes deviation between i) medium-and short-term schedules and ii) production goals. We correspondingly present novel techniques to improve the model tractability. Constraints such as precedence, mill and extraction capacities, heat, and diesel emissions are considered. The model produces a consistent schedule while ensuring the safety of the work environment. TE19 CC Room 203A In Person: Healthcare General Session Chair: Zilong Wang, Atlanta, GA, 30318, United States 1 - An Analysis of Incentive Schemes for Participant Retention in Clinical Studies Xueze Song, University of Illinois-Urbana-Champaign, Champaign, IL, United States, Mili Mehrotra, Tharanga K. Rajapakshe Patient retention is one of the critical issues that haunt clinical studies. This paper analyzes two interventions—incentive payment and effort—and identify their optimal combination for retaining desired number of participants. Also, we examine several commonly observed payment schemes in practice and compare their relative performances under different settings. 2 - Regret Analysis for Adaptive Model Predictive Control Ilgin Dogan, Ph.D. Candidate, University of California, Berkeley, Berkeley, CA, 94707-2017, United States, Zuo-Jun Max Shen, Anil Aswani The exploration/exploitation trade-off is an inherent challenge in data-driven and adaptive control. Though this trade-off has been studied for multi-armed bandits, reinforcement learning for finite Markov chains and linear control systems; it is less well-studied for adaptive control of nonlinear systems. A significant challenge in nonlinear setting is that there is no explicit characterization of an optimal policy for a given set of parameters. We propose a new regret notion with respect to a finite-horizon oracle controller with full knowledge of parameters and develop a new class of learning-based policies in the context of adaptive model predictive control. We conduct statistical analyses to prove finite sample concentration bounds for the estimation step, and then we perform theoretical analyses to show the closed-loop stability and low regret of our policy. 3 - Estimating Heterogeneous Treatment Effects with Modern Mixed Integer Programming Formulations The recent data driven approach paradigm at identifying and solving problems has reignited interest amongst policy makers, businesses, and physicians in shifting away from a “one-size-fits-all” mentality to a more tailored approach. Estimating heterogeneous treatment effects (HTEs) is thus of great import in many fields such as personalized medicine, marketing, and policy evaluation. State of the art techniques, are either based on highly restrictive classifiers or too complex to interpret. In this paper, we unify and generalize these approaches under a Generalized Method of Moments (GMM) estimation framework, and formulate it as a Mixed Integer Program (MIP). This estimation framework presents a “best-of-both-worlds” approach which simultaneously achieves interpretability and flexibility in classification problems based on treatment effect heterogeneity. Zilong Wang, Georgia Institute of Technology, Atlanta, GA, United States, Zhaowei She, Turgay Ayer, Shihao Yang

TE17 CC Room 202A In Person: Game Theory Contributed Session Chair: Krista J. Li, Indiana University, Bloomington, IN, 47405, United States 1 - Social Welfare Maximization and Conformism via Information Design in Linear-quadratic-gaussian Games Furkan Sezer, PhD Student, Texas A&M University- College Station, College Station, TX, United States, Hossein Khazaei, Ceyhun Eksin We consider linear-quadratic-gaussian (LQG) games in which players have quadratic payoffs that depend on the players’ actions and an unknown payoff- relevant state, and signals on the state that follow a Gaussian distribution conditional on the state realization. An information designer decides the fidelity of information revealed to the players in order to maximize the social welfare of the players or reduce the disagreement among players’ actions. Leveraging the semi-definiteness of the information design problem, we derive analytical solutions for these objectives under specific LQG games. 2 - Illegal Fishing in Congested Maritime Environments Michael M. Perry, George Mason University Volgenau School of Information Technology and Engineering, Fairfax, VA, United States A maritime environment is modeled where two countries in close proximity must delineate fishing rights. Countries issue fishing quotas and its shown one can benefit significantly from issuing excessive quotas, inducing illegal fishing in the other’s legal waters. The costs imposed by patrol craft serve as a way of deterring this behavior, but patrols can be offset by employing martial assets onboard fishing vessels, a phenomenon of increasing regularity. The cost-effectiveness of each of these measures is assessed. The paper concludes that even if the countries can agree to cooperative combat illegal fishing, the potentiality of illegal fishing will significantly alter the terms of such a deal. 3 - A Proof of the Optimality of the E-patrolling Strategy for the Continuous Patrolling Game Thuy Bui, Rutgers Business School, Newark, NJ, United States, Thomas Lidbetter We consider the continuous patrolling game introduced in Alpern et al. (2016). This is a zero-sum game between an Attacker, who attacks a network at a particular time and place, and a Patroller, who patrols the network with the aim of intercepting the attack. Recently, Alpern et al. (2020) conjectured that a patrolling strategy called the E-patrolling strategy is optimal for all tree networks, and they proved this to be true in some special cases. In this paper, we settle the conjecture by providing -optimal strategies for the Attacker. We also give precisely optimal Attacker strategies for some previously unsolved star networks. 4 - Product Innovation with Customer Recognition Krista J. Li, Associate Professor, Indiana University, Bloomington, IN, United States We examine how customer recognition affects brands’ incentives to invest in product innovation. We find that when brands have similar equity, customer recognition increases brands’ incentives to invest in product innovation. However, when brands have sufficiently different equity, customer recognition leads the stronger brand to invest more and the weaker brand to invest less in product innovation. In addition, customer recognition can increase the weaker brand’s profit but decreases it more for the stronger brand. Thus, collecting customers’ purchase history data for customer recognition can be beneficial for weaker brands but detrimental for stronger brands. TE18 CC Room 202B In Person: Advances in Mining General Session Chair: Alexandra M Newman, Colorado School of Mines, Golden, CO, 80401-1887, United States 1 - Characterizing Heat and Diesel Emissions in an Underground Mine for a Production Scheduling Model Aaron Swift, BS, Colorado School of Mines, Golden, CO, 80401, United States Current production scheduling models for underground mines do not consider real-time heat or emissions from diesel equipment, which can lead to unsafe conditions as heat or emissions accumulate in the working areas. Based on thermodynamic principles, this research develops a transient model for heat, emissions, and ventilation that can be infused into a short-term production scheduling model for near-term operational planning.

142