2015 Informs Annual Meeting

WE21

INFORMS Philadelphia – 2015

WE21 21-Franklin 11, Marriott Mathematical Modeling of HIV at the Cellular, Individual, and Population Level Sponsor: Health Applications Sponsored Session Chair: Eva Enns, Assistant Professor, Univeristy of Minnesota, 420 Delaware St. SE, Minneapolis, MN, 55408, United States of America, eenns@umn.edu 1 - Mathematical Modeling of HIV Viral Dynamics and Immune Response During Treatment Interruptions Nargesalsadat Dorratoltaj, Virginia Tech, Room 392, Vet Med Phase II., Blacksburg, VA, 24060, United States of America, nargesd@vt.edu, Josep Bassaganya-riera, Stanca Ciupe, Stephen Eubank, Margaret O’dell, Hazhir Rahmandad, Kaja Abbas The objective of this study is to understand and predict HIV viral and immune dynamics at the individual level during treatment interruption and low adherence in developing AIDS. We use ordinary differential equations to build immune-viral dynamics of HIV/AIDS during treatment interruption. The results of the simulation show that the time that patients start treatment and for how long they stay on treatment before interruption are unique predictor for time to progress to AIDS. 2 - A Data-driven Approach to Understanding HIV Care Trajectories and Retention in Care Challenges Eva Enns, Assistant Professor, Univeristy of Minnesota, Sub-optimal retention in HIV primary care is associated with increased risks of mortality, detectable viral loads, and other complications compared to regular, ongoing care; however, the standard dichotomous measures of retention (e.g., in/out of care) ignore the diversity and complexity of individual care trajectories. In this talk, I describe a data-driven approach to understanding HIV care patterns and how to incorporate complex care behaviors into mathematical models of HIV interventions. 3 - Modeling a Bi-layer Contact Network of Injection Drug Users and The Spread of Blood-borne Infections Rui Fu, Stanford University, 74 Barnes CT, Stanford, CA, 94305, United States of America, ruif@stanford.edu, Alexander Gutfraind, Margaret L. Brandeau Blood-borne infections can spread among injection drug users (IDUs) via risky sexual and needle-sharing contacts. To accurately model the spread of such contagions among IDUs, we build a bi-layer network that captures both types of risky contacts. We present methodology for inferring important model parameters. Such a model can be used to evaluate the efficacy of various programs that aim to combat drug addiction and contain blood-borne diseases among IDUs. 4 - Measuring the Potential Impact of Combination Prevention in Sub-Saharan Africa Amin Khademi, Assistant Professor, Clemson University, 130-D Freeman Hall, Clemson, SC, 29634, United States of America, khademi@clemson.edu, Sunanth Anand A public health approach to combination HIV prevention is advocated to contain the epidemic in Sub-Saharan Africa. We explore the implications of universal access to treatment along with HIV education scale-up in the region. We develop an HIV transmission model, calibrate it with data from South Africa and simulate the impacts of universal access to treatment along with HIV education scale-up on HIV trends. 420 Delaware St. SE, Minneapolis, MN, 55408, United States of America, eenns@umn.edu

2 - Transportation Optimizer System as Applied to Rail Car Assignment Problem Ilya Buzytsky, BIAS Intelligence, 7741 37th Ave NE, Seattle, WA, 98115, United States of America, ilya@biasintelligence.com BIAS Optimizer System was developed to address problems frequently encountered in many transportation and distribution logistics applications. Its objective is to find optimal solution to problems that can be formulated in terms of supply/cost/demand configuration across large matrix representing a multi- node network: The solution helps find an assignment that would meet the demand supported by available supply across all nodes and satisfy all business rules and constraints at the minimum cost. 3 - Diet Problem Pevisited Fariborz Partovi, Professor, Drexel University, 33rd Chestnut Street, Department of Decision Sciences, Philadlphia, PA, 19003, United States of America, Partovi@Drexel.edu It has been close to seventy years since the Diet problem was introduced by Stigler (Stigler 1945). However the problem with the classical Diet models is based on lack of proper presentation of food preferences. For many people, especially when they are eating outside their home, the content of the food as far as nutrition is concern may be not as important as the taste of the food. In this article we modify Diet problem using extensions of linear programming to consider the above issues

WE19 19-Franklin 9, Marriott Computational Optimization with Risk and Uncertainty Sponsor: Computing Society Sponsored Session

Chair: John Siirola, Sandia National Laboratories, P.O. Box 5800, MS 1326, Albuquerque, NM, 87185-1326, United States of America, jdsiiro@sandia.gov 1 - Multistage Power Generation Capacity Expansion Models with Different Risk Measures Shu Tu, Lehigh University, H.S. Mohler Laboratory, 200 West Packer, Bethlehem, PA, 18015, United States of America, sht213@lehigh.edu, Boris Defourny We investigate different stochastic optimization formulations for the multistage power generation capacity expansion problem. In particular, we focus on risk John Siirola, Sandia National Laboratories, P.O. Box 5800, MS 1326, Albuquerque, NM, 87185-1326, United States of America, jdsiiro@sandia.gov, Jean-paul Watson, David Woodruff Progressive Hedging (PH) is a scalable and effective approach for solving large stochastic programming problems through scenario-based decomposition. However, PH is sensitive to key tuning parameters (notably rho) and for many problems can exhibit slow convergence. In this work we present new approaches for accelerating convergence and improved tuning by propagating key information among scenarios. 3 - Production Planning under Uncertainty and Service Level Constraints Suleyman Karabuk, Associate Professor, University of Oklahoma, School of Industrial and Sys Engineering, 202 W. Boyd St., Room 124, Norman, OK, 73019, United States of America, karabuk@ou.edu Consider a production planning problem with machine changeover and inventory carrying costs, and service level constraints, where demand for products is probabilistic. We formulate the problem as a multistage stochastic programming model with recourse. The resulting model is very large scale even for trivially small size instances. We develop an effective decomposition algorithm where individual product sub problems are solved by a novel probabilistic dynamic programming model. measures whose parameters can be calibrated from market data. 2 - Exploiting Inter-scenario Information to Acclerate Progressive Hedging

485