Informs Annual Meeting 2017

MD71

INFORMS Houston – 2017

3 - Employee or Contractor: A Study of Courts’ Misclassification Decisions

3 - Computational Approaches to Mixed Integer Second Order Cone Optimization Aykut Bulut, Lehigh University, Bethlehem, PA, United States, aykut@lehigh.edu, Ted Ralphs We introduce a framework for solving Mixed Integer Second Order Cone optimization Problems (MISOCP). We present solver DisCO implemented within COIN-OR’s CHiPPS framework. DisCO is a flexible library that implements different algorithms to solve MISOCP. 4 - Cutting Plane Techniques via Decision Diagrams Danial Davarnia, Carnegie Mellon University, ddavarni@andrew.cmu.edu, Willem-Jan Van Hoeve We develop cutting plane techniques for certain classes of discrete problems based on their decision diagram formulation. For linear IPs, these cutting planes provide a natural generalization of Gomory and intersection cuts. For convex IPs, they dominate the linearized cuts used in the outer approximation schemes. These cutting planes can also be derived for nonconvex IPs which allows for an adaptation of outer approximation algorithms. Computational experiments show significant gap improvement upon the traditional cutting plane methods.

Mohammad Javad Feizollahi, Assistant Professor, UW-Madison, Georgia State University, Atlanta, GA, 30303, United States, mfeizollahi@gsu.edu, Charlotte Alexander In this research, we study federal district court decisions in worker misclassification cases, 2008-2016. Independent contractors lack the benefits and protections available to employees. They may file suit, seeking “employee” classification, but the law poorly defines the two categories. Moreover, judges’ classification opinions have never been studied systematically. We use text mining and analytical tools to understand how courts distinguish between employees and independent contractors and the factors that influence their decisions. 4 - Margin-based Pattern Selection for Support Vector Machines Dongil Kim, Korea Institute of Industrial Technology, 89 Yangdaegiro-gil, Ipjang-myeon, Seobuk-gu, Cheonan, Korea, Republic of, dikim01@kitech.re.kr, Seokho Kang, Sungzoon Cho Support Vector Machines (SVM) is a widely used training algorithm in data mining and machine learning. However, as the size of the dataset increases, the training complexity of the SVM has become a limit to apply SVM to real applications. In this paper, we propose a margin-based pattern selection method for SVM. Margin of each training pattern is estimated from multiple SVMs trained with smaller sample sets. Based on the estimated margin, the proposed method selects patterns that are likely to be support vectors. Experiments conducted on various datasets show that the proposed method has good generalization performance. 5 - Data Mining for Individual Consumer Credit Default Prediction under E-commence Context: A Comparative Study Jilei Zhou, Guanghua School of Management, Peking University, No.5 Yiheyuan Road Haidian District, Beijing, 100871, China, 1501110963@pku.edu.cn This paper explored the relative effectiveness of alternative classifiers (decision tree, logistic regression, random forests and extreme gradient boosting) to estimate how likely an applicant is to default in individual consumer credit services offered by e-commerce platform. Specifically, our work contributes to the following research questions: (i) What kind of features should be introduced in the new context of e-commence? Which features plays important roles in credit scoring model? (ii) How to tuning classification algorithms to avoid model inefficiency? (iii) Do ensemble classifiers real improve classification ability? Data used came from an financial services platform in China. 371F Mixed-Integer Polynomial Optimization Sponsored: Optimization, Global Optimization Sponsored Session Chair: Akshay Gupte, Clemson University, Clemson, SC, 29634-0975, United States, agupte@clemson.edu 1 - Outer-product-free Sets for Polynomial Optimization and Oracle-based Cuts Chen Chen, Columbia University, chen.chen@columbia.edu, Daniel Bienstock, Gonzalo Munoz We consider valid inequalities for the intersection of a closed set S with a polyhedron P. Given an oracle providing the distance to S we propose a pure cutting plane algorithm; if the initial relaxation is a polytope, the algorithm converges. Cuts are generated from convex forbidden zones, or S-free sets derived from the oracle. A special case is polynomial optimization represented with a symmetric matrix of decision variables. In this lifted representation we let S be the set of real, symmetric outer products, and accordingly develop a theory of outer-product-free sets. We identify two families of maximal outer-product-free sets; associated intersection cuts can be generated in polynomial time. 2 - On Branching-point Selection for Triple Products in Spatial Branch-and-bound: the Hull Relaxation Emily Speakman, University of Michigan, 2222 Fuller Court, Apt 702A, Ann Arbor, MI, 48105, United States, eespeakm@umich.edu, Jon Lee Having three or more expressions multiplied together occurs frequently in factorable global-optimization models. For these triple products, we present some analytic results regarding the choice of branching variable and branching point in the context of sBB for factorable formulations. Following the approach we previously introduced to compare various common relaxations associated with trilinear monomials, we use 4-dimensional volume as a comparison measure to evaluate the sBB choice of branching variable and branching point in this context. For branching-point selection, we also compare our results to common practice in software. MD71

MD72

372A Networks and Supply Chains Sponsored: Optimization, Network Optimization Sponsored Session

Chair: Anna B Nagurney, University of Massachusetts-Amherst, University of Massachusetts-Amherst, Amherst, MA, 01003, United States, nagurney@isenberg.umass.edu 1 - A Closed-loop Supply Chain Equilibrium Model with Random and Price-sensitive Demand and Return Qiang Qiang, Penn State, Management Division, 30 E Swedesford Rd, Malvern, PA, 19355, United States, qzq10@psu.edu, Younes Hamdouch This paper proposes a decentralized closed-loop supply chain network model consisting of raw material suppliers, manufacturers, retailers, and recovery centers. The demands for the product and the corresponding returns are random and price-sensitive. Retailers and recovery centers face penalties associated with shortage demand and supply, respectively.The optimality conditions of the various decision-makers are derived. The governing equilibrium conditions can be formulated as a finite-dimensional variational inequality problem. The qualitative properties of the solution to the variational inequality are discussed. Numerical examples are provided for illustration purpose. 2 - Supply Chain Network Capacity Competition with Outsourcing: A Variational Equilibrium Framework Deniz Besik, Master’s, Isenberg School of Management, University of Massachusetts Amherst, Amherst, MA, 01003, United States, dbesik@som.umass.edu, Anna Nagurney, Min Yu This paper develops a supply chain network game theory framework with multiple manufacturers/producers, with multiple manufacturing plants, who own distribution centers, which are distinguished by brands, to demand markets, while maximizing profits and competing noncooperatively. The manufacturers avail themselves of external distribution centers for storing their products and freight service provision. We impose capacities on the links. We utilize a special case of the Generalized Nash Equilibrium problem, known as a variational equilibrium. A case study on apple farmers in Massachusetts is provided to illustrate the potential benefits of outsourcing in this sector. 3 - Mergers and Acquisitions in Blood Banking Systems: A Supply Chain Network Approach Amir H. Masoumi, Assistant Professor of Management, Manhattan College, School of Business, De La Salle Hall 508, Riverdale, NY, 10471, United States, amir.masoumi@manhattan.edu, Min Yu, Anna B.Nagurney We develop a methodological framework for the assessment of total cost efficiency (synergy) associated with a merger or acquisition in the blood banking industry, which is experiencing a volatile environment. The network optimization pre- and post-merger models handle perishability of the life-saving product of blood, include both operational and discarding costs of waste, capture the uncertainty associated with the demand points, as well as the expected total blood supply shortage cost and the total discarding cost at demand points. The models incorporate capacities on the links and their solution yields the optimal link flows plus frequencies of activities throughout the blood supply chain.

262