2015 Informs Annual Meeting

SB19

INFORMS Philadelphia – 2015

SB17 17-Franklin 7, Marriott Networks Robustness and Vulnerability Analysis Sponsor: Optimization/Network Optimization Sponsored Session Chair: Foad Mahdavi Pajouh, Assistant Professor, University of Massachusetts Boston, 100 Morrissey Boulevard, Boston, MA, 02125, United States of America, Foad.Mahdavi@umb.edu 1 - On Imposing Connectivity Constraints in Integer Programs Austin Buchanan, Oklahoma State University, 322 Engineering In many network applications, one searches for a connected subset of vertices that exhibits other desirable properties. To this end, we study the connected subgraph polytope of a graph, which is the convex hull of subsets of vertices that induce a connected subgraph. We investigate two classes of valid inequalities—- separator inequalities and indegree inequalities—-and show when they induce (all nontrivial) facets. We also consider extended formulations, lifting, and separation. 2 - On Biconnected and Fragile Subgraphs of Low Diameter Oleksandra Yezerska, Texas A&M University, 3131 TAMU, College Station, TX, 77843, United States of America, yaleksa@tamu.edu, Sergiy Butenko, Foad Mahdavi Pajouh An s-club is a subset of vertices inducing a subgraph with a diameter of at most s. It is commonly used to characterize network clusters in applications for which easy reachability between group members is of high importance. In this paper, we study two special cases of the 2-club model – a biconnected 2-club, and a fragile (not biconnected) 2-club, respectively. By investigating certain properties of both models, we develop exact algorithms for their corresponding optimization problems. 3 - Integer Programming Formulations for Solving the Minimum Edge Blocker Spanning Tree Problem Jose Walteros, University at Buffalo, 342 Bell Hall, Buffalo, NY, United States of America, josewalt@buffalo.edu, Ningji Wei, Foad Mahdavi Pajouh The minimum edge blocker spanning tree problem consists of removing a minimum number of edges in a weighted graph so that the weight of any spanning tree in the remaining graph is at least r. We study the convex hull of feasible solutions and identify facet-inducing inequalities for this polytope. We develop an exact algorithm that solves our formulation via branch-and-cut. Finally, we provide the computational results obtained after solving a set of randomly generated and real-life instances. Joint Session Modeling Methodologies/Big Data: Large-Scale Data Analytics and Applications Cluster: Modeling and Methodologies in Big Data Invited Session Chair: Chou-An Chou, Binghamton University, 4400 Vestal Parkway, Vestal, United States of America, cachou@binghamton.edu Co-Chair: Anas Hourani, Binghamton University, 4400 Vestal Parkway, Vestal, United States of America, ahouran1@binghamton.edu 1 - Understanding Patterns or Relations of the Terrorist Attacks in Big Data to Prevent Future Threats Salih Tutun, Binghamton University, 4400 Vestal Parkway East, Binghamton, NY, 13902, United States of America, stutun1@binghamton.edu, Chou-An Chou This research interests value in 5Vs that is the useful results about terrorist attacks (incidents) to analyze the terrorist activity patterns or relations, to predict their future moves, and finally to prevent potential terrorist attacks. We focused on understanding of why incident succeed or fail, duration of incident extended, and doubt as to whether the incident is an act of terrorism. The results will be very useful for law enforcement agencies to propose reactive strategies. North, Stillwater, OK, 74078, United States of America, buchanan@okstate.edu, Sergiy Butenko, Yiming Wang SB18 18-Franklin 8, Marriott

2 - Machine Learning Based Robust Optimization with Application to Healthcare Treatment Recommendations Sung Hoon Chung, Binghamton University, P.O. Box 6000, Binghamton, NY, United States of America, schung@binghamton.edu, Yinglei Li When making decisions, one may use robust optimization (RO) to reduce the impact of uncertainty modeled from the past data. In RO, uncertain parameters are generally assumed to be within a convex set, within which decision makers want to protect the system against the worst case. We propose a machine learning approach to design such uncertainty sets, and discuss the application of machine learning based robust optimization to healthcare treatment recommendations. 3 - Frequency-based Rule Classification Algorithm with Big Data Anas Hourani, Binghamton University, 4400 Vestal Parkway, Vestal, NY, United States of America, ahouran1@binghamton.edu, Chou-An Chou Frequency-based rule classification algorithm is a simple, fast and accurate associative algorithm. In this study, we are going to introduce an improvement on the FRC algorithm to work Hadoop system. Then several comparisons are made between the conventional FRC and FRC on Hadoop to demonstrate the efficiency of the proposed algorithm with big data. SB19 19-Franklin 9, Marriott Computational Optimization for Applied Problems I Sponsor: Computing Society Sponsored Session Chair: David Morrison, Inverse Limit, 255 McDowell Lane Unit A, W Sacramento, CA, 95605, United States of America, dmorrison@invlim.com 1 - Enumeration and Bounding Arguments for Infrastructure Resilience Analysis W. Matthew Carlyle, Naval Postgraduate School, mcarlyle@nps.edu, David Alderson We propose a functional definition of infrastructure resilience based on attacker- defender models and using a parametric analysis of attacker capability. In general, this parametric analysis requires enumeration of a potentially enormous number of optimization problems. In this talk, we present a computational technique that uses bounding arguments to significantly limit the enumeration while still providing useful measures of infrastructure resilience. We illustrate with a real case study. 2 - Computational Study of Stabilization Methods in Crew Pairing Problems Jiadong Wang, Senior Operations Research Developer, Sabre, 3150 Sabre Drive, 76092, United States of America, jiadwang@gmail.com, Xiaodong Luo Column generation has been successfully applied to solve large-scale optimization problems. In this talk, we review the various stabilization methods by revisiting primal-dual subproblem approach. Experiments on large set-partitioning instances from crew pairing problem in Sabre provide insight on the effectiveness of various stabilization methods. 3 - An Implicit Hitting-set Solution Method for the Sensor Location Problem David Morrison, Inverse Limit, 255 McDowell Lane Unit A, W Sacramento, CA, 95605, United States of America, dmorrison@invlim.com, Paul Rubin Given a two-way directed network, the sensor location problem seeks to find the minimal number of vertices that must be monitored so that flow over the entire network can be determined, given prior assumptions on flow patterns. We reduce the problem to an implicit hitting set problem and use Benders’ decomposition to solve the problem. The decomposition subproblem is a novel network flow problem called the incremental flow problem, for which we present several solution techniques.

71