2016 INFORMS Annual Meeting Program

SD86

INFORMS Nashville – 2016

SD86 GIbson Board Room-Omni

SD91 Davidson Ballroom A-MCC Joint Session HAS/MSOM-HC: Statistical Decision- Making with Applications in Healthcare Sponsored: Manufacturing & Service Oper Mgmt, Healthcare Operations Sponsored Session Chair: Mohsen Bayati, Stanford University, Stanford, CA, United States, bayati@stanford.edu Co-Chair: Hamsa Bastani, Stanford University, 10 Comstock Circle, Apt 304, Stanford, CA, 94305, United States, hsridhar@stanford.edu 1 - Approximation Methods For Adaptive Clinical Trial Design John R Birge, University of Chicago, John.Birge@ChicagoBooth.edu 2 - An Analytics Approach To Designing Drug Therapies For Cancer John M Silberholz, MIT, Cambridge, MA, 02139, United States, josilber@mit.edu, Dimitris Bertsimas We present a data-driven approach to planning clinical trials and designing novel drug therapies for metastatic breast cancer (MBC). First, we describe construction of a large database of MBC clinical trial results and tools to help clinicians visualize the data. Next, we use statistical models to predict efficacy and toxicity outcomes of trials before they are run, with implications for selecting between multiple drug therapies for testing. Finally, we use optimization models to design novel therapies that strike a balance between maximizing patient outcomes and learning about new drugs; initial evaluation suggests these models may improve trial outcomes compared to current practice. 3 - Online Decision-making With High-dimensional Covariates Hamsa Sridhar Bastani, Stanford University, 10 Comstock Circle, Apt 304, Stanford, CA, 94305, United States, hsridhar@stanford.edu, Mohsen Bayati Big data has enabled decision-makers to personalize choices based on an individual’s observed characteristics. We formulate this problem as a multi-armed bandit with high-dimensional covariates, and present a new efficient algorithm that provably achieves near-optimal performance. The key step in our analysis is proving convergence of the LASSO estimator despite non-iid data induced by the bandit policy. We evaluate our algorithm using a real patient dataset on warfarin dosing; here, a patient’s optimal dosage depends on her genetic profile and medical records. Our algorithm outperforms existing bandit methods as well as physicians to correctly dose a majority of patients. 4 - Estimating Average Treatment Effects In High-dimensional Observational Studies Stefan Wager, Stanford University, Stanford, CA, United States, swager@stanford.edu, Susan Athey, Guido Imbens There are many studies where researchers are interested in estimating average treatment effects and are willing to rely on the unconfoundedness assumption, which requires that treatment assignment is as good as random conditional on pre-treatment variables. The unconfoundedness assumption is often more plausible if a large number of pre-treatment variables are included in the analysis, but this can worsen the finite sample properties of existing approaches to estimation. In this paper, we propose a new method for estimating average treatment effects in high dimensions that achieves the semi-parametric efficiency bound without requiring any modeling assumptions on the propensity score. SD92 Davidson Ballroom B-MCC INFORMS Optimization Society Prize Session Award Session Chair: Suvrajeet Sen, University of Southern California, 3715 McClintock Ave, Los Angeles, CA, 90089, United States, s.sen@usc.edu 1 - Optimization SocietyAwards Suvrajeet Sen, University of Southern California, 3715 McClintock Ave, Los Angeles, CA, 90089, United States, s.sen@usc.edu The Optimization Society sponsors four awards annually. They are a) the Khachiyan Prize for lifetime contributions in optimization, b) the Farkas Prize for exceptional mid-career accomplishments, c) the Young Optimization Researcher award, and finally, d) the student paper prize competition. These awards are highly competitive and coveted, and this session is dedicated to congratulating the winners, and their lasting contributions to optimization. The award winners will present brief overviews of their prize-winning contributions.

Manufacturing IV Contributed Session

Chair: Ali AlArjani, PhD Candidate, University of Wisconsin - Milwaukee, 4848 N. Lydell Ave, Apt 141, Milwaukee, WI, 53217, United States, alarjan2@uwm.edu 1 - Setting Optimal Planned Leadtimes In A Configure To Order Manufacturing System Sjors Jansen, PhD Candidate, Eindhoven University of Technology (TU/e), P.O. Box 513, Paviljoen E13, Eindhoven, 5600MB, Netherlands, s.w.f.jansen@tue.nl, Zumbul Atan, Ton de Kok, Ivo Adan We study the production planning in Configure To Order (CTO) manufacturing systems. The system consists of multiple stages that converge to one final assembly stage. Leadtimes per stage are stochastic due to extensive testing at the end of each stage. Our goal is to determine optimal planned leadtimes for each stage such that the total expected production costs are minimized. We derive Newsvendor equations for each individual stage. This set of equations is solved and the exact optimal planned leadtimes for each stage are obtained. These equations give an important insight in the dynamics of the system, since they indicate to what extend a specific stage can be blamed for the lateness of the final product. 2 - Introduction Of A Motor Assembly Test Bed To Verify Manufacturing Technology Jungryul Bae, Korea Institute of Industrial Technology (KITECH), Seoul, Korea, Republic of, somanythat@naver.com, SeHwan Ahn, YongJu Cho, Chul Kim, Hyunchul Tae A Testbed is a place of verifying newly developed manufacturing technologies before apply it in practice. We built a motor-assembly Testbed that comprises seven connected facilities in a line. We have tested several manufacturing technologies including simulation, quality management, and IoTs on the Testbed. In this presentation, we aim to introduce our Testbed and share our experience. The final goal of the Testbed is to enhance the localization ratio of a convergence of the IoTs and manufacturing technology.Keywords: Connected Smart Factory (CSF), Testbed, IoTs, manufacturing technology. 3 - Modeling The Impact Of Product Variety On Inventory: Application To Strategic Assembly Sequencing And Supply Chain Design Jeonghan Ko, University of Michigan; Ajou University, 1205 Beal Ave., Industrial & Operations Engineering, Ann Arbor, MI, 4810, United States, jeonghan@umich.edu, Heng Kuang This paper models the impact of variety on assembly supply chain design when limited commonality exists between products. We derive theorems on the impact of product variety on safety inventory, and provide a measure to approximate the impact. The theorems and new measure are applied in two problems: optimal process sequencing and optimal assembly decomposition. We prove that to prioritize the process with a smaller number of variants will reduce the supply chain cost no matter the commonality is. 4 - Finite Capacity Material Requirement Planning System For Supply Chain Network Benyaphorn Paopongchuang, Sirindhorn International Institute of Technology, Pathum Thani, 12121, Thailand, bp322@njit.edu Benyaphorn Paopongchuang, New Jersey Institute of Technology, Newark, NJ, 07102, United States, bp322@njit.edu, Pisal Yenradee Available Finite Capacity Material Requirement Planning (FCMRP) systems have some limitations. They are designed to determine production and purchasing plans in only one factory not in a multi-level supply chain network. Most systems lack of optimization capabilities. In addition, they do not manage bottleneck effectively. The proposed algorithm tries to develop FCMRP system that also considers finite capacity of some key suppliers and customers. 5 - Similarity Coefficient Model For Solving an Oil Global Facility Location Problem Ali AlArjani, PhD Candidate, University of Wisconsin - Milwaukee, 4848 N lydell Ave, Milwaukee, WI, 53217, United States, alarjan2@uwm.edu Solve an oil global facility location problem by a new similarity coefficient model for cluster analysis that model ranks multiple countries and cluster them in groups each group have similar attributes.

120