Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MC01

16 - Evaluation of Primary Care Delivery with E-visits Aditya Mahadev Prakash, University of Florida, 473 Weil Hall, Gainesville, FL, 32608, United States, Xiang Zhong E-visit is a messaging service offered by physicians through patient portals. Offering e-visit entails a horizontal substitution of office visits for a segment of patients. A key problem for an institution that provides both services is to optimally allocate resources that improves patients’ care-access, profitably. A queuing model with non-homogenous arrival rates is developed. Service capacity regions that improve accessibility and conditions for profitability are identified. Under high traffic, based on patient segmentation, medical institutions can equip their e-visit services with an optimal amount of resources to ensure an economical outcome while providing care to more patients. 17 - Optimal Integration of Desensitization Protocols into Kidney Paired Donation Programs Fatemeh Karami, University of Louisville, Louisville, KY, 40292, United States, Mehdi Nayebpour, Monica Gentili, Naoru Koizumi, Keith Melancon Blood type (ABO) incompatibility and antibody to donor human leukocyte antigen (HLA) remain the most significant barriers in transplantation. While pre- transplant desensitization can be administered to overcome such incompatibilities between living donors and their kidney recipients, desensitization alone is likely to fail for those pairs with significant incompatibilities. For these pairs, desensitization can be administered in combination with Kidney Paired Donor (KPD) exchange, the system that allows incompatible pairs to exchange donors with other incompatible pairs to improve donor-recipient compatibilities. 18 - Spare Parts Estimation Using Cox Modeling Alejandro Najera-Acosta, Graduate Student, New Mexico State University,Las Cruces, NM, 88001, United States, Delia J. Valles- Rosales In a competitive production environment, only those companies that consider all aspects of processes or systems performance remain competitive. Management of spare components is a key feature for the performance of maintenance activities. Spare parts constitute an essential element in all industries, they are designed for a specific use, its useful life is random, and its propagation is difficult to determine. Therefore, spare parts inventories are established to allow rapid replacement of failed parts and ensure a continuity of the operations. An approach through Cox Model is proposed for its estimation to improve accuracy in the inventory management of spare parts. 19 - Optimal Placement of Actuators for Composite Fuselage Shape Control Juan Du, Peking University, College of Engineering, Beijing, 100871, China, Xiaowei Yue, Jeffrey H. Hunt, Jianjun Shi Actuator placement is critical and challenging for shape control due to dimensional variabilities of composite fuselages. Current practice is non-optimal and low efficient. We propose an optimal actuator placement methodology for efficient composite fuselage shape control by developing a sparse learning model and corresponding estimation algorithm. The case study shows that our proposed method achieves the optimal actuator placement for shape adjustments of the composite fuselage. n MC01 North Bldg 121A Computationally Tractable Methods for Stochastic Optimization Sponsored: Optimization/Optimization Under Uncertainty Sponsored Session Chair: Sarah M. Ryan, Iowa State University, Ames, IA, 50011-2164, United States 1 - New Approaches for Solving Two-stage Stochastic Mixed-integer Programs with Continuous Recourse Siavash Tabrizian, Southern Methodist University, Dallas, TX, 75206, United States, Harsha Gangammanavar, Halit Uster In this talk we present enhancements to the L-shaped method to solve large-scale two-stage stochastic mixed-integer programs with continuous recourse. We apply sampling techniques within optimization to achieve computational improvement. We demonstrate these results on classical problems in the literature. 2 - Robust Optimization for Linear Gaussian Processes Georgios Kotsalis, ISyE Ga Tech, Atlanta, GA, 30309-5413, United States, Guanghui Lan We provide a computationally tractable procedure of affine policies for the constrainedmultistage robust optimization problem as it pertains to linear models that are subject to quadratic constraints, while being affected by uncertain Monday, 1:30PM - 3:00PM

external disturbances and Gaussian noise. We derive our results under the general assumption that the external disturbanceslie within some nominal range expressed as the intersection of ellipsoids centered at the origin. A particular class of problems that falls under the scope of our investigations are mass- transportation problems requiring the optimal steering of a linear stochastic system to a finite probability distribution. 3 - Observational Data-Based Quality Assessment of Scenario Generation for Stochastic Programs Sarah M. Ryan, Iowa State University, 3004 Black Engineering, Industrial & Manufacturing Systems Eng, Ames, IA, 50011-2164, United States, Didem Sari Ay The quality of stochastic program solutions depends on the quality of scenarios employed to obtain them. Given a set of historical instances, an appealing way to assess scenario generation is to conduct a backtest of the scenario generation and solution procedure, in which the cost of the optimized first-stage solution is assessed using the observed values of the uncertain parameters. Such a study may be very demanding computationally. We propose alternative approaches using past instances that do not require solving deterministic equivalents. Instead, we assess the quality of scenario sets by applying reliability metrics to the optimal costs of single-scenario subproblems. n MC02 North Bldg 121B Robust Optimization Sponsored: Optimization/Optimization Under Uncertainty Sponsored Session Chair: Michael R. Wagner, University of Washington,Seattle, WA, 98195, United States 1 - Robust Monitoring on Social Networks with a Application to Suicide Prevention Aida Rahmattalabi, University of Southern California, Los Angeles, CA, United States We consider the problem of selecting “gatekeepers,” with uncertain availabilities, to train as monitors capable of recognizing warning signs of suicide among their peers in a social network. We formulate the problem as a two-stage robust optimization problem that aims to maximize the worst-case number of covered nodes. We propose a practically tractable approximation scheme based on the K- adaptability idea. We demonstrate the effectiveness of our proposed approach on various network instances. In particular, we perform a case study on a real social network of college students. Finally, we illustrate how our solution can be used to inform gatekeeper training on college campuses. 2 - Pricing Service Level Guarantees in Cloud Computing Chaithanya Bandi, Kellogg School of Management, Northwestern University, 2211 Campus Dr, Room 4169, Evanston, IL, 60208, United States We consider the problem of pricing service level agreements in Cloud computing systems. We leverage tools from Robust Queueing theory and multi-stage optimization to compute the prices. 3 - A Robust Multi-period Newsvendor Model with Inventory Balance Constraints Saumya Sinha, University of Washington, Box 353925, University of Washington, Seattle, WA, 98195, United States, Michael R. Wagner, Archis Ghate We present a robust newsvendor model that accounts for revenue in addition to the various cost parameters, thereby leading to a profit maximization problem under inventory balance constraints. We provide closed-form expressions for the optimal order quantities. This calls for the solution of the so-called inner problems, which are also solved analytically for a large class of commonly occurring uncertainty sets. 4 - A Robust Optimization Approach to Crowdsourcing Last-mile Deliveries Soraya Fatehi, University of Washington, Michael G. Foster School of Business, University of Washington, 358 Mackenzie Hall, Seattle, WA, 98195-3200, United States We propose and analyze a robust optimization model for crowdsourcing last-mile deliveries via independent drivers. In this context, our model minimizes the worst-case delivery cost of a firm, subject to delivering all customer orders on- time, with high probability. We derive the optimal proportion of packages to assign to the crowd, the optimal number of packages per driver, the optimal crowdsourced delivery area, and the optimal crowd workforce. We show that, if the crowd-delivery system is designed optimally, then the firm can significantly benefit from crowdsourcing last-mile deliveries, especially for fast same-day deliveries.

184