2015 Informs Annual Meeting

SD02

INFORMS Philadelphia – 2015

SD02 02-Room 302, Marriott OR and Homeland Security 1: Data Driven Decisions Cluster: Homeland Security Invited Session Chair: Paul Kantor, Prof, Rutgers, 96 Frelinghuysen Dr, Piscataway, NJ, United States of America, paul.kantor@rutgers.edu 1 - Detecting and Locating GPS Jamming Jeff Coffed, Senior Marketing Manager, Exelis Inc., 400 Initiative Drive, Rochester, NY, 14606, United States of America, Jeffrey.Coffed@exelisinc.com, Joe Rolli GPS has become a ubiquitous service supporting critical infrastructure. Its signal is susceptible to service blockages due to jamming. Recognizing this threat, Exelis set out to develop technology that identifies and locates jamming sources. The system can be located around high-risk areas to instantaneously sense and locate jamming sources. Users will receive pin-point geolocation information in order to respond. We will share information about the threat, the technology and test results. 2 - Walk through Metal Detectors at Sports Stadiums Christie Nelson, Postdoctoral Associate, CCICADA, Rutgers University, 96 Frelinghuysen Rd, 4th Floor, CoRE Building, Piscataway, NJ, United States of America, Christie.L.Nelson.PhD@gmail.com, Paul Kantor, Fred Roberts, Dennis Egan, Brian Ricks, Michael Tobia, Brian Nakamura, Ryan Whytlaw, Michael Young Experimental designs are presented for walk through metal detectors (WTMDs) in stadium settings. Experiments were created to understand how WTMDs perform in real settings, typically outdoors, as opposed to idealized indoor lab scenarios. Experiments were then carried out at sports stadiums. Because of the large number of experimental factors involved, a combinatorial experimental design approach was taken. 3 - A Differential Privacy Mechanism for Graph Problems Protecting Confidential Network Data William Pottenger, Rutgers University, 96 Frelinghuysen Road, CoRE Building, Piscataway, NJ, 08854, United States of America, billp@dimacs.rutgers.edu, Kunikazu Yoda Graph problems are important for homeland security since most critical infrastructure such as a power grid can be modeled by a graph. Such network data often contains highly sensitive information and a publically released summary must not give hints to terrorists who might exploit the data in developing targets. We present a differential privacy mechanism for graph problems whose solutions reveal useful global information while not revealing significant confidential individual information. 4 - Fusion Learning by Individual-to-Clique (FLIC): Efficient Approach to Enhancing Individual Inference Minge Xie, Professor, Rutgers University, 501 Hill Center, Piscataway, United States of America, mxie@stat.rutgers.edu Learning from multiple studies can often be fused together to yield a more effective inference. We present a new approach, named “Fusion Learning by Individual-to-Clique (FLIC)”, to enhancing inference of an individual study through adaptive combination of confidence distributions obtained from its clique (namely similar studies). Drawing inference from the clique allows borrowing strength from similar studies to enhance the inference. It can also substantially reduce computational expense. 5 - The Unaccompanied Alien Children Challenge: Applying Queuing Theory to Improve Logistics in Immigration The recent wave of unaccompanied alien children that crossed the border of the United States posed a very important logistic challenge for the US DHS and US DHHS. In this presentation, Queuing Theory is presented as an option to forecast the performance of facilities aimed at carrying pre-screenings of these immigrants before their final placement in shelters. We present a mathematically tractable queuing model, illustrate its capabilities, and discuss opportunities that this approach offers. Javier Rubio-Herrero, Rutgers University, Piscataway, NJ, United States of America, javier.rubioherrero@rutgers.edu

SD03 03-Room 303, Marriott

Application of Scheduling Theory Cluster: Scheduling and Project Management Invited Session

Chair: Zhixin Liu, University of Michigan-Dearborn, 19000 Hubbard Drive, Dearborn, MI, United States of America, zhixin@umich.edu 1 - Optimization Based Production Scheduling with Batching Lixin Tang, Professor, Northeastern University, Institute of Industrial Engineering and, Logistics Optimization, Shenyang, 110819, China, lixintang@mail.neu.edu.cn We discuss the production scheduling problems with batching decision arising from steel, petrochemical and non-ferrous metal industry. For general problems, complexity and optimal solution properties are analyzed; polynomial time algorithm for solvable cases, and approximation algorithm with theoretical analysis for NP-hard problems are proposed. For complicated problems, row- column generation algorithm, LR&CG based dual algorithm, and improved Benders decomposition algorithm are proposed. 2 - An Optimization Model for Loan Collection Ping He, Zhejiang University, 866 Yuhangtang Rd, Hangzhou, China, phe@zju.edu.cn, Zhongsheng Hua, Zhixin Liu This paper studies how to make efficient collection decisions over consumer term- loan accounts. Since a loan’s onset, an account experiences state transition across ages. We model the state transition of loan accounts using a Markov transition matrix, and provide optimization method to determine the collection action at each state and age for each consumer type that maximizes the lender’s expected value. Managerial insights and general rules for consumer loan collection are recommended. 3 - Two-Agent Scheduling on a Single P-Batching Machine with Equal Processing Time and Non-Identical Job Jun-Qiang Wang, Professor, Northwestern Polytechnical University, Box 554, No. 127 West Youyi Road, Department of Industrial Engineering, Xi’an, 710072, China, wangjq@nwpu.edu.cn, Cheng-wu Zhang, Yingqian Zhang, Guo-qiang Fan, Joseph Leung We schedule two agents on a single parallel-batching machine. For the linear weighted sum model, we presented an approximation algorithm and analyze the absolute/asymptotic worst-case ratio. For the restriction model, no approximation algorithm with a finite bound exists, unless P = NP. We propose two polynomial- time heuristic algorithms using two restriction-solving strategies. For the non-domination model, we define a boundary of Pareto-optimal set. The proposed heuristics outperform NSGA-II. 4 - Vessel and Containing Planning in Feeder Lines Yu Wang, PhD Candidate, Dept. of IELM, HKUST, Rm 5567, Academic Building, Clear Water Bay, Kowloon, 999077, Hong Kong - PRC, ywangbi@connect.ust.hk, Xiangtong Qi We consider the vessel and container planning problem in feeder lines. A feeder vessel sequentially visits n ports, collecting containers from each port and transporting to hub port. The route of the vessel is pre-defined. The optimal serving policy for the vessel to load and unload under stochastic demand is investigated. The process is described as a Markov decision process, which aims to maximize the expected revenue. Furthermore, the optimal loading and unloading policy is derived. SD04 04-Room 304, Marriott Special Panel on 20th Year Anniversary of WORMS: Strategies for Advancing Women in OR/MS Sponsor: Women in OR/MS Sponsored Session Chair: Guzin Bayraksan, Associate Professor, The Ohio State University, Integrated Systems Engineering, Columbus, OH, 43209, United States of America, bayraksan.1@osu.edu 1 - Special Panel on 20th Year Anniversary of WORMS: Strategies for Advancing Women in OR/MS Moderator:Guzin Bayraksan, Associate Professor, The Ohio State University, Integrated Systems Engineering, Columbus, OH, 43209, United States of America, bayraksan.1@osu.edu, Panelists: Laura Mclay, Margarit Khachatryan, Paula Lipka, Candace Yano This special panel will look back at the 20 years of Women in OR/MS. Then, it will focus on strategies for advancing women in OR/MS. Topics include how to recruit and retain women students and faculty in Operations Research and Management Science, academic and industrial leadership.

120