Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

POSTER SESSION

3 - Identifying and Visualizing Uncommon Customer Response on Machine Learning Seung Wan Seo, Korea University, Seoul, Korea, Republic of,, Deokseong Seo, Myungjun Jang, Jaeyun Jung, Pilsung Kang Detecting and visualizing uncommon but notable opinions from a large amount of Voice of Customers (VOC) data, is an important procedure in response to customer complaints. In this study, we propose a framework for identifying uncommon but significant responses and frequently used words in them based on distributed document representation, local outlier factor, and TF-IDF methods. We also propose a result visualization procedure that can provide useful information to vehicle engineers. This uncommon response detection and visualization framework can accelerate the efficiency and effectiveness of a large amount of VOC data analytics. 4 - Deep Learning Based Damage Detection on Post-hurricane Satellite Imagery Quoc Dung Cao, University of Washington, Seattle, WA, 98105- 0002, United States After a hurricane, damage assessment is critical to emergency managers and first responders. To improve the efficiency and accuracy of damage assessment, instead of using windshield survey, we propose to automatically detect damaged buildings using image classification algorithms. The method is applied to the case study of 2017 Hurricane Harvey. 5 - A Systematic Architecture Definition Methodology for Engineering Prognostic and Health Management System Rui Li, Delft University of Technology, Delft, Netherlands, Wim J.C. Verhagen, Richard Curran Prognostics and health management (PHM) has become a crucial research in many engineering fields, enabling fault diagnostics, prognostics remaining useful lifetime and health management. However, the characteristics of PHM architecture methodology have not been dealt with in depth. To fill the gaps, this research defines a systematic architecture definition methodology for engineering a PHM system. 6 - Optimal Inventory Distribution Strategy for Relief Supply Considering Information Uncertainty after a Major Disaster Riki Kawase, Kobe University, Nada, Kobe, Hyogo, Japan Humanitarian supply chain management plays an important role in addressing major disasters. However, timely supply was impossible because of mismatching of the needs and the bottleneck of the last mile. Now, two types of plans have been proposed: (1) öDirect distribution and (2) öKanban systemö, but no mathematical analysis has been done. Our research conducts mathematical analysis on these empirical proposals. Specifically, we find the optimal strategy of the inventory distribution problem using the optimal control theory. We analyze the optimal control and clarify the effectiveness of proposal (1). Additionally, numerical example shows that updating information can make the system worse. 7 - Optimal Switching Sequence for Switched Linear Systems Zeyang Wu, University of Minnesota, Minneapolis, MN, 55414- 2922, United States, Qie He We study a mixed-integer nonlinear optimization problem. This problem has many applications in operations research and control, yet a moderate-sized instance is challenging to solve to optimality for state-of-the-art optimization software. We propose a simple exact algorithm for this problem. Our algorithm runs in polynomial time when oligo-vertex property, a concept we introduce in this paper for a set of matrices, is satisfied. We derive several sufficient conditions for a set of matrices to have the oligo-vertex property. Numerical results demonstrate the clear advantage of our algorithm in solving large-sized instances of the problem over one state-of-the-art global solver. 8 - Integrating Vehicle Routing and Scheduling to Optimize Foster Care Visitations Caroline M. Johnston, Worcester Polytechnic Institute, Worcester, MA, 01609, United States, Shima Azizi, Katherine Dunphy, Renata Konrad, Andrew C. Trapp We have partnered with a county in New York State that oversees approximately 100 foster care cases, each requiring the transportation of children to weekly/bi- weekly meetings with biological parents and case workers at an assigned meeting time. The current driver/case worker assignment system is executed manually in a cumbersome manner. We model this problem using combinatorial optimization concepts such as the Dial-A-Ride and team orienteering problems with the goal of maximizing the throughput of cases seen per week. Our goal is to create a decision-support tool to recommend weekly driver/case worker assignments to facilitate this scheduling process, increasing the quality of this system.

9 - Virus Spread Over Networks Philip Pare, University of Illinois at Urbana Champaign, Urbana, IL, 61801, United States, Carolyn Beck, Angelia Nedich, Tamer Basar Virus models over non-trivial networks are commonly motivated by biological, computer, and human contact networks. Developing spread models over networks that can be analyzed and validated can help develop efficient mitigation techniques. Previous study of spread models has been done, however, learning the spread parameters of such models has not yet been explored, and the models have not been validated by real data. We present several analysis results, employ John Snow’s seminal work on cholera epidemics in London in the 1850’s to validate a network-dependent susceptible-infected-susceptible (SIS) model, and present a control heuristic for mitigating the spread of an epidemic. 10 - The Reliable Facility Location Model Considering Dependent Supporting Structure Prakash Jamar Kattel, Ohio University, Athens, OH, 45701, United States, Felipe Aros-Vera This research develops a mathematical optimization-based framework to determine the optimal location of a critical facility in interconnected networks. This is important since the failure of a single facility in a network might produce cascading failures in interdependent Critical infrastructures and produce negative social and economic impacts. 11 - Optimization of Food Pantry Locations to Address Food Scarcity in Toledo Ece Sanci, University of Michigan, Ann Arbor, MI, 48105, United States, Sharanya Chandran, Jie Ma, Emily Morris, Jeremy Pasteris Food for Thought (FFT) is a nonprofit organization working in downtown Toledo and surrounding areas. FFT works to alleviate food scarcity among food insecure households. In this study, we propose an optimization model to help FFT determine optimal timing and locations for their pantry deployment. 12 - Mixed Strategy Nash Equilibrium (NE) for Flight-frequencies Competition in Airlines Market Chun-Han Wang, National Tsing Hua University, Hsinchu, Taiwan Yu-Ching Lee, Yue Dai In airlines market, each company competes on flight-frequencies and airport time slots to pursue higher profits. We construct equilibrium programming models to compute the exact NE flight-frequencies, including basic model where each airline acts as a player and alliance model where each alliance acts as a player. It is known that there may not exist a pure strategy NE. Therefore, mixed strategy NE is formulated instead. We anticipate to get existence result, expected total profits for each player, and a suggested percentage of code-share flights of each airline within an alliance. 13 - Learning and Analysis of Precedence Network Based on Coupon Subset Collection Problem Xiaotian Xie, University of Illinois at Urbana Champaign, Urbana, IL, 61801, United States, Vincent Hoff, Carolyn Beck We infer manufacturing precedence relationships using process data, where sequenced subassembly tasks are viewed as nodes, and edges represent precedence constraints between tasks. Given subassembly begin and end times, we identify precedence relationships to analyze differences in planned versus implemented processes. Contributions include inference algorithms and corresponding sample complexity analyses. 14 - Niche Centrality and Compression in Competitive Organizational Networks Brian Aronson, Duke University, Durham, NC, 27705, United States Organizations compete across many dimensions and at varying intensities, resulting in complex structures of competition. Prior work finds that organizations’ survival chances are influenced by population dynamics across broadly-defined organizational niches; however, less is known about patterns of competition within such niches. This paper constructs a new model for studying competitive organizational networks and hypothesizes two structural mechanisms that influence organizations’ survival chances: Niche centrality and compression. 15 - Equitable Pricing of Episodes of Care in a Cluster-based Bundled Payment System Bikram Partap Singh, M.S. Candidate, Rochester Institute of Technology, Rochester, NY, 14623, United States, Ruben A. Proaño This study proposes a systemic approach to simultaneously price multiple episodes of care under a cluster-based bundle payment framework. We present a multi- criteria optimization model that makes highly expensive episodes of care more affordable by reallocating revenue expectations among less expensive clusters of encounters for different episodes of care.

183