2015 Informs Annual Meeting

TA19

INFORMS Philadelphia – 2015

4 - Integrated Variable Importance Assessment in Multi-stage Manufacturing Processes Gianluca Gazzola, Rutgers Center for Operations Research, 100 Rockafeller Road, Piscataway, NJ, 08854, United States of America, ggazzola@scarletmail.rutgers.edu, Jeongsub Choi, Myong K (MK) Jeong, Byunghoon Kim We introduce a method for the assessment of variable importance in manufacturing processes characterized by a hierarchy of technical relationships between stage variables. Regression models of direct technical relationships and a novel permutation measure are employed to quantify the local contribution of every variable. Global contributions are finally obtained by integrating these local assessments, based on the overall structure of indirect and direct technical relationships in the process. TA19 19-Franklin 9, Marriott Computational Integer Optimization Sponsor: Computing Society Sponsored Session Chair: Yan Xu, Director, SAS, 100 SAS Campus Dr., Cary, NC, United States of America, yan.xu@sas.com 1 - Recent Advances in the FICO Xpress MIP Solver Michael Perregaard, Xpress Team, FICO, International Square, Starley Way, Birmingham, B37 7GN, United Kingdom, MichaelPerregaard@fico.com We will present some of the recent MIP advances in the FICO Xpress solver, with an emphasis on how it is able to exploit the ever increasing core counts of modern CPUs. 2 - The SAS MILP Solver: Current Status and Future Developments Philipp Christophel, SAS Institute Inc., 100 SAS Campus Dr., Cary, NC, 27607, United States of America, Philipp.Christophel@sas.com, Menal Guzelsoy, Imre Polik, Amar Narisetty We give an overview of the current status of the SAS mixed integer linear programming (MILP) solver that is part of the SAS/OR product. The focus will be on describing recent implementation efforts for the MILP presolver as well as future development directions. 3 - Performance Improvements and New Features in the Gurobi Optimizer Chris Maes, Senior Developer, Gurobi Optimization, Inc., 125 Beacon St, Apt. #4, Boston, MA, 02116, United States of America, maes@gurobi.com This talk will cover the latest developments in the Gurobi Optimizer. We’ll discuss the new Gurobi Cloud, which makes it easy to launch one or more Gurobi machines when you need them. We’ll also talk about our upcoming release, which includes significant performance enhancements and several new features. 4 - CPLEX Keeps Getting Better Andrea Tramontani, CPLEX Optimization, IBM Italy, Via Martin Luther King 38/2, Bologna, Italy, andrea.tramontani@it.ibm.com We present some of the new features and algorithmic techniques that have been recently added to IBM ILOG CPLEX Optimizer, and we give detailed benchmark results that demonstrate the performance improvements achieved in latest CPLEX versions.

information, or computational limitations. We recover missing interactions from vast corporate networks using novel biclique clustering techniques to detect the most significant edges, and hence provide new insights into power structures. 2 - Running Your Optimization Model on the Cloud with the IBM CPLEX Studio IDE Frederic Delhoume, Software Engineer, IBM, 9 Rue de Verdun, Gentilly, 94253, France, delhoume@fr.ibm.com We will show how to easily run optimization models from the IBM CPLEX Studio IDE. We will also demonstrate how to monitor the cloud service and get local results from the remote optimization service. A REST API way of running models on the cloud will be shown. TA21 21-Franklin 11, Marriott Medical Decision Making in Cancer Care Sponsor: Health Applications Sponsored Session Chair: Christine Barnett, University of Michigan, 1205 Beal Ave., Ann Arbor, MI, United States of America, clbarnet@umich.edu 1 - Predictive Modeling for Optimal Design of Cancer Detection Protocols Selin Merdan, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States of America, smerdan@umich.edu, Brian Denton Diagnosis of chronic diseases often involves expensive and invasive tests and procedures. Predictive models can play an important role in determining the optimal diagnostic protocol based on individual patient risk factors. We discuss an approach for developing predictive models using clinical observational data that suffers from common sources of bias such as low disease prevalence and missing data. We illustrate the use of these models for optimization of prostate cancer diagnostic protocols. 2 - Model-based Calibration for Natural History Modeling Jing Voon Chen, University of Southern California, Epstein Dept of Indus & Sys Eng, Los Angeles, CA, United States of America, jingvooc@usc.edu, Julie Higle A natural history (NH) model often requires calibration of unobservable model parameters to fit observed data. Uncertainty in the data and in the calibrated parameters impacts confidence in the optimal decision. We propose a method for model-based calibration that is resilient to these uncertainties, especially for comparative analyses of disease screening or treatment strategies. Illustrative examples and sensitivity analyses will be discussed. 3 - Assessment of Individualized Human Papillomavirus (HPV) Vaccination Strategies Fan Wang, University of Arkansas, 4207 Bell Engineering Center, The human papillomavirus (HPV) is the most common sexually transmitted virus in the U.S. To prevent multiple cancers attributable to the HPV, HPV vaccine is recommended for preteens and teens who have not been exposed to HPV. We develop a simulation model for the optimal design of personalized HPV vaccination program, which incorporates multiple social-behavioral and demographic risk factors. The efficacy of the HPV vaccination program is evaluated in terms of the HPV-related health outcomes. 4 - Tailoring CRC Screening Strategy for Different Age- and Gender-specific Population Subgroups Carolina Vivas, Purdue University, West Lafayette, IN 47906, United States of America, cvivas@purdue.edu, Nan Kong, Robert Klein, Thomas Imperiale Standard guidelines for colorectal cancer (CRC) strategies do not consider different age- and gender-specific subgroups for tailored screening recommendations. Recent evidence suggests that men tend to face a higher risk of developing advance adenomas earlier than women. We apply Design of Experiments techniques to quantify the risk differences on CRC disease progression. Model based cost-effectiveness analyses of various screening strategies are conducted for different population subgroups. Fayetteville, AR, United States of America, fxw005@email.uark.edu, Shengfan Zhang

TA20 20-Franklin 10, Marriott Cloud Services and Applications Cluster: Cloud Computing Invited Session

Chair: Grace Lin, Data Analytic Technology and Applications (DATA), Data Analytic Technology and Applications (DATA), Taipei, Taiwan - ROC, gracelin@iii.org.tw 1 - Revealing Power Structures through Novel Biclustering Approaches Sabine Baumann, Professor Dr., Jade University, College of Mgmt, Info., Tech., Friedrich-Paffrath-Str. 101, Wilhelmshaven, 22880, Germany, sabine.baumann@jade-hs.de, Oliver Eulenstein, Christoph Wunck Cloud and big data provide unprecedented access to massive interaction networks of people and organizations. However, exploring such rich data environments encounters equally extensive challenges: unreliable, incomplete or distorted

264