2015 Informs Annual Meeting

SC17

INFORMS Philadelphia – 2015

SC17 17-Franklin 7, Marriott Network Optimization Sponsor: Optimization/Network Optimization Sponsored Session Chair: Kelly Sullivan, Assistant Professor, University of Arkansas, Fayetteville, AR, 72701, ksulliv@uark.edu 1 - A Decomposition Approach for Dynamic Network Interdiction Models Chase Rainwater, University of Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, United States of America, cer@uark.edu, Forough Enayaty Ahangar This work details the development of a large-scale optimization approach for solving dynamic bilevel network interdiction problems. A benders decomposition approach is proposed that utilizes constraint programming to exploit the scheduling nature of the network interdiction subproblem solved over a finite time horizon. Computational results comparing the proposed approach to traditional constraint programming and mixed-integer programming approaches are discussed. 2 - Supply Chain Design through Acquisition: A Robust Multi-objective Approach Amin Khademi, Assistant Professor, Clemson University, 130-D Freeman Hall, Clemson University, Clemson, SC, 29634, United States of America, khademi@clemson.edu, Mariah Magagnotti, Scott Mason Combining supply chain networks for an acquisition is a complicated process; making decisions for strategic merging of the supply chains when working with incomplete or incorrect data is even more so. Our work presents a robust optimization model that allows the decision maker to adjust for an expected degree of uncertainty, thus producing solutions that are less responsive to incorrect or incomplete data without being excessively cautious. 3 - Multiple-scenario Approach for a Dynamic Disaster Relief Routing Problem with Uncertain Social Data Emre Kirac, PhD Candidate, University of Arkansas, Department of Industrial Engineering, Fayetteville, AR, 72701, United States of America, ekirac@email.uark.edu, Ashlea Milburn Social media may play an important role and improve situational awareness in disaster response by providing real-time information. We present decision support models capable of considering input streams from social data when planning for disaster response. Specifically, a dynamic routing problem is presented in which social data and information from trusted sources are available and change over time. Alternative decision policies are presented and compared across a variety of request scenarios. 4 - The Wireless Network Jamming Problem Subject to Protocol Interference Hugh Medal, Mississippi State University, Industrial & Systems Engineering, Starkville, United States of America, hugh.medal@msstate.edu We study a wireless network jamming problem, solving it using a cutting plane approach that is able to solve networks with up to 81 transmitters. Our study yields the following insights into wireless network jamming: 1) increasing the number of channels is the best strategy for designing a robust network, and 2) increasing the jammer range is the best strategy for the attacker.

our model not only dramatically reduce the number of selected features and control the costs of the features, but also has a promising accuracy in medical diagnosis. 2 - Networked Data Classification with Node Selection Daehan Won, University of Washington, Seattle, 1415 NE We present a framework for classification of networked structure data . Due to the huge size of the network, we apply a feature selection scheme. Instead of general feature selection methods, we present a node selection scheme to determine the most relevant sub-networks which might yield insightful information underlying complex networks. 3 - A Structural Model and Bayesian Estimation of New Technology Adoption Sebastian Souyris, PhD Candidate, The University of Texas at Austin, 2110 Speedway Stop B6500, Austin, TX, 78712, United States of America, sebastian.souyris@utexas.edu, Jason A. Duan, Anant Balakrishnan, Varun Rai We present a structural model and estimation algorithm to analyze the adoption of a new technology at individual consumer level. The model incorporates networks, e.g. geographical distance, to estimate the potential effect of word of mouth by assuming that the previous adoptions affect the predisposition of a consumer towards the technology. For inference, we use a Bayesian algorithm that overcomes the computational burden of classical estimation methods of structural models. Ravenna Blvd, #401, Seattle, WA, 98105, United States of America, wondae@uw.edu Chair: Jingyang Xu, The Walt Disney Company, 1375 East Buena Vista Drive, Orlando, FL, 32830, United States of America, jxu7@buffalo.edu 1 - Consolidation and Last-mile Costs Reduction in Intermodal Transport Martijn Mes, University of Twente, P.O. Box 217, Enschede, Netherlands, m.r.k.mes@utwente.nl, Arturo Pérez Rivera We consider a carrier that transports freight periodically, using long-haul round trips from a single origin to multiple last-mile locations, and vice versa. Since the long-hauls are always traveled, the last-mile locations determine the costs of each trip. The challenge is to select, for each trip, the combination of orders which reduces costs over time. We propose an approximate dynamic programming (ADP) approach, which we illustrate using data from a Dutch intermodal carrier. 2 - A Real-time Run-curve Computation Framework for Trains with Dynamic Travel Restrictions Jingyang Xu, The Walt Disney Company, 1375 East Buena Vista Drive, Orlando, FL, 32830, United States of America, jxu7@buffalo.edu, Daniel Nikovski, Sae Kimura We study the problem to generate the most energy efficient run-curves subject to given travel time requirements and speed limit changes. We propose a two stage procedure framework. With derived geometric relations, the actual run-curves are generated in the real-time stage using approximate dynamic programming. Computational results show that the framework is capable to generate near- optimal run-curves in real time. 3 - Emission Oriented Multi-objective Sensor Location Model on Freeway Ning Zhu, Assistant Professor, Tianjin University, Weijin Road, No. 92, Tianjin, China, zhuning@tju.edu.cn, Shoufeng Ma, Qinxiao Yu, Yuche Chen In our study, an interpolation method is proposed to reconstruct the vehicle trajectory on a second-by-second base by using traffic sensors. A multi-objective traffic sensor location model is proposed aiming to estimate four major pollutants accordingly. Different numerical experiments are conducted in various freeway topological structures. It shows that the interpolated method for estimating vehicle trajectory can have a reasonable good emission accuracy. SC19 19-Franklin 9, Marriott Application in Transportation Systems Sponsor: Computing Society Sponsored Session

SC18 18-Franklin 8, Marriott Data Mining for Healthcare Cluster: Modeling and Methodologies in Big Data Invited Session

Chair: Daehan Won, University of Washington, Seattle, 1415 NE Ravenna Blvd, #401, Seattle, WA, 98105, United States of America, wondae@uw.edu 1 - General Framework for Rulebased Medical Diagnosis and Decision Making Chunyan (sally) Duan, Tongji University & University of Washington, A503, Sino-French Center, Tongji University,

No.1239, Siping Road, Shanghai, 200092, China, duanchunyan87@gmail.com, Daehan Won, Ying Lin, Shuai Huang, Jianxin You, W. Art Chaovalitwongse

A new mixed integer programming model is developed to select rules with minimization of prediction errors and control the costs of the features in the framework. The Diabetes Prevention Trial-Type 1 dataset is used to evaluate and compare our model with other State-of-the-Art methods. The results show that

98