Informs Annual Meeting 2017

MB71

INFORMS Houston – 2017

MB72

6 - Mortality Risk Prediction for Atrial Fibrillation Patients Safak Yakti, PhD Candidate, Binghamton University, Binghamton, NY, United States, syakti1@binghamton.edu, Mohammad T.Khasawneh, Serkan Saygi Atrial Fibrillation (AF) is the most common heart rhythm disorder worldwide according to CDC. AF affects more than 3 million Americans and causes 130,000 death per year. Aim of this research is to identify the factors that influence mortality caused by AF and to compare between the prediction models namely Logistic Regression, Decision Trees, and Alternating Decision Tree methods. Performance measures are used to evaluate performances of the models. These models are utilized by using two different datasets; first one imbalanced dataset (7,796 patients), and second under-sampled dataset (2,350 patients) with AF diagnosis who were referred between 2012 and 2014. 371F MINLP Methods and Applications Sponsored: Optimization, Global Optimization Sponsored Session Chair: Christos Maravelias, PhD, University of Wisconsin - Madison, Madison, WI, United States, christos.maravelias@wisc.edu 1 - Integer Programming Approaches for Appointment Scheduling with Random No-shows and Service Durations Ruiwei Jiang, University of Michigan, 1205 Beal Ave., Ann Arbor, MI, 48109, United States, ruiwei@umich.edu, Siqian Shen, Yiling Zhang We consider an appointment scheduling problem with random no-show and service duration. We propose a distributionally robust optimization model and derive computationally tractable reformulations based on integer programming approaches. Numerical results demonstrate that the proposed approach yields better out -of-sample performance, especially when the randomness is misspecified. 2 - Closing the Gap NNAC Optimal Power Flow by Strengthening the Convex QC Relaxation using Parallel Computing Gabriel Hackebeil, hackebeg@umich.edu An important problem in power systems is finding optimal operating conditions for transmission systems, which are governed by the non-convex AC power flow equations (AC-OPF). In this work, we apply global optimization strategies, such as optimality-based range reduction and spatial branch-and-bound, in an HPC setting to close the optimality gap on many of the AC-OPF test cases from NESTA. We use the convex QC relaxation as a subproblem and employ linearization techniques to avoid numerical issues with solving this SOCP in practice, allowing for faster and more robust subproblem solves. Our implementation is built on the open-source AML Pyomo. 3 - An MINLP Model and Solution Methods for Blend Scheduling Problems Christos T. Maravelias, University of Wisconsin-Madison, 1415 Engineering Drive, Engineering Hall 2024, Madison, WI, 53706, United States, maravelias@wisc.edu, Yifu Chen We propose a new mixed integer nonlinear programming (MINLP) model for integrated blending and scheduling problems in oil refineries. Unlike existing approaches, the model considers crude oil operations, intermediate processing and final product blending simultaneously. To solve the proposed model efficiently, we develop preprocessing techniques based on propagation of product demands and property specifications, and propose three families of tightening constrains. We present computational results showing the effectiveness of these methods. 4 - Exploiting Integrality in the Global Optimization of Mixed-integer Nonlinear Programming Problems with Baron Nikolaos Sahinidis, Swearingen Professor, Carnegie Mellon University, Department of Chemical Engineering, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, sahinidis@cmu.edu, Mustafa Kilinc We present recent developments in the global optimization software BARON to address problems with integer variables. A primary development was the addition of mixed-integer linear programming relaxations to BARON’s portfolio of linear and nonlinear programming relaxations. We describe BARON’s dynamic strategy for deciding under what conditions to activate integer programming relaxations in the course of branch-and-bound. We also describe primal heuristics for finding good solutions of mixed-integer nonlinear programs. Finally, we report extensive computational results to analyze the impact of each technique in the solution process. MB71

372A Network Optimization VI Sponsored: Optimization, Network Optimization Sponsored Session

Chair: Foad Mahdavi Pajouh, University of Massachusetts Boston, Boston, MA, 02125-3393, United States, mahdavi@okstate.edu 1 - Finding Central Clique Clusters in Graphs Foad Mahdavi Pajouh, University of Massachusetts Boston, Boston, MA, Foad.Mahdavi@umb.edu, Maciej Rysz We propose a solution algorithm for identifying the most central cliques in graphs and examine its effectiveness when the centrality measure is defined by betweenness. Numerical experiments demonstrating the computational performance of the proposed method are conducted and compared with results obtained from solving an equivalent mixed integer programming representation. 2 - Extreme-point Search Methods for Solving Fixed-charge Generalized Networks Angelika Leskovskaya, Southern Methodist University, Caruth Hall 3145 Dyer Street, Suite 372, Dallas, TX, 75275, United States, aleskovs@lyle.smu.edu, Richard Barr While researchers have studied generalized networks flow problems extensively, the powerful addition of fixed charges on arcs has received scant attention. This work describes network-simplex-based algorithms that efficiently exploit the quasi-tree basis structure of the problem relaxations and presents computational comparisons with commercial solution alternatives. 3 - Finding Groups with Maximum Betweenness Centrality Oleg Prokopyev, University of Pittsburgh, Pittsburgh, PA, 15261, United States, droleg@pitt.edu, Alexander Veremyev In this talk we consider the problem of identifying the most influential (or central) group of nodes (of some predefined size) in a network. Such a group has the largest value of betweenness centrality or one of its variants, for example, the length-scaled or the bounded-distance betweenness centralities. We demonstrate that this problem can be modelled as a mixed integer program (MIP) that can be solved for reasonably sized network instances using off-the-shelf MIP solvers. We also present extensive computational experiments with different types of real-life and synthetic network instances to illustrate the effectiveness and flexibility of the proposed framework. 372B Tutorial on Meta-algorithms Sponsored: Computing Sponsored Session Chair: Meinolf Sellmann, IBM, 1101 Kitchawan Rd, Yorktown Heights, NY, 10, United States, meinolf@us.ibm.com 1 - Meta-Algorithms Meinolf Sellmann, IBM, 1101 Kitchawan Rd, Yorktown Heights, NY, United States, meinolf@us.ibm.com Meta-algorithmics is a subject on the intersection of learning and optimization whose objective is the development of effective automatic tools that tune algorithm parameters and, at runtime, choose the approach that is best suited for the given input. In this tutorial, we summarize the core lessons learned when devising such meta-algorithmic tools. MB73

190