2015 Informs Annual Meeting

MA38

INFORMS Philadelphia – 2015

2 - New Strategies for Quantifying the Resilience of Supply Chains to Temporally Distinct Disruptions Jacqueline Griffin, Assistant Professor, Northeastern University, 334 Snell Engineering Center, 360 Huntington Ave, Boston, MA, 02125, United States of America, ja.griffin@neu.edu, Ozlem Ergun, Shiqing Liu The saline supply chain network flow formulation applied for a multi-level supply chain with lead time between each level and concerning about how factors would influence each other in different time periods. We present closed form expressions to characterize the resilience of a supply chain network to varying combinations of temporally distinct disruptions. 3 - Exploring Strategies for Private Sector Transportation in Uganda Jarrod Goentzel, MIT, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States of America, goentzel@mit.edu, Mark Brennan, Emily Gooding New product technology is commonly introduced in developing countries through subsidized pilot programs run by non-governmental organizations (NGOs). Low landed cost is key for further scaling up product distribution through the private sector. This study uses a pilot program for agricultural storage products in Uganda to explore strategies to reduce transportation cost. 4 - Tracking Healthcare Associated Infections at Individual Level over Dynamic Human Networks Ziye Zhou, The Chinese University of Hong Kong, William M W Healthcare associated infections (HAIs) have become a major challenge to public healthcare. This work addresses the problem of tracking the transmission of HAIs at an individual level. We present a framework with three key components of time-varying contact network construction, individual-level transmission tracking and HAI parameter estimation. Experiments on human positioning data collected in a four-month tracking study in a hospital are conducted to evaluate the performance. MA37 37-Room 414, Marriott Health Care Modeling and Optimization VI Contributed Session Chair: Md Noor E Alam, Post Doctoral Fellow, Massachusetts Institute of Technology, 135 Quincy Ave, Apt 204, Quincy, MA, 02169, United States of America, mnalam@mit.edu 1 - Shift Scheduling for an Anesthesiology Residency Program Hernan Abeledo, Associate Professor, George Washington University, 800 22 St. NW, Washington, DC, 20052, United States of America, abeledo@gwu.edu, Michael Kanter, Ian Morgan, Jean - Max Buteau, Liam Nealon Creating shift schedules for resident physicians is a notoriously difficult task that is typically done manually by the chief residents. Shift assignments need to observe a large number of rules, as well as adhere to fairness and desirability factors while populating a very complex schedule structure. We present an integer programming model developed to schedule anesthesiology residents at the George Washington University Hospital. 2 - Shift Scheduling for Medical Residency Programs Anthony Coudert, George Washington University, 800 22 St. NW, Washington, DC, United States of America, coudert@gmail.com, Hernan Abeledo Creating shift schedules for resident physicians is a tedious task that is typically done manually by the chief residents. Shift assignments need to observe a large number of rules, as well as adhere to fairness and desirability factors while populating a very complex schedule structure. We present integer programming models used to schedule residency programs at the George Washington University Hospital. 3 - Open-access Outpatient Clinic Scheduling Yu Fu, ISEN Dept. Texas A&M University, 3131 TAMU, College Station, TX, 77843, United States of America, yufu@tamu.edu, Amarnath Banerjee This study aims at exploring cost-efficient offline and online scheduling methods under the open access policy which allows the visits of the same-day-request patients and walk-in patients as compensation for no-shows of regular patients to improve clinic performance and revenue benefit. The offline scheduling uses approximation and heuristic methods on scenarios and data generated by prediction and simulation. The online scheduling relies on heuristic methods and stochastic programming models. Mong Engineering Bldg., Hong Kong, Hong Kong - PRC, zhouzy@se.cuhk.edu.hk, Chun-hung Cheng, Dobin Ng

4 - Integer Linear Programming Based Statistical Techniques for Causal Inference Md Noor E Alam, Post Doctoral Fellow, Massachusetts Institute of Technology, 135 Quincy Ave, Apt. 204, Quincy, MA, 02169, United States of America, mnalam@mit.edu, Cynthia Rudin Organizations are fiercely struggling to realize valuable information from large- scale data that are increasingly used for understanding important cause and effect relationships. This research developed a methodological frameworks to solve such critical problems with ILP based statistical techniques. One of the key idea is to develop robust techniques to handle uncertainty in data driven decision making, particularly as applied to healthcare. MA38 38-Room 415, Marriott Applied Probability I Contributed Session Chair: Giang Trinh Senior Research Associate, Ehrenberg-Bass Institute, University of South Australia, 70 North Terrace, Adelaide, SA, Australia, giang.trinh@marketingscience.info 1 - Value of Communication in a One-leader, Two-followers Partially Observed Markov Game Yanling Chang, PhD Candidate, Georgia Institute of Technology, 765 Ferst Dr, Atlanta, GA, 30332, United States of America, changyanling@gatech.edu, Alan Erera, Chelsea White We consider a one-leader, two-followers partially observed Markov game and analyze how the value of the leader’s criterion changes due to changes in the communication quality between the two followers. We present conditions under which the value of the leader’s criterion degrades or improves, as a function of this communication quality and the type of game (collaborative or non- collaborative). 2 - Multi-period Corporate Survival Probability Estimation with Stochastic Covariates Ahmad Reza Pourghaderi, Assistant Professor, Abdullah Gul University, Department of Industrial Engineering, Melikgazi, Kayseri, 38039, Turkey, pourghaderi@u.nus.edu, Ebrahim Sadreddin We propose an econometric method to obtain maximum likelihood estimation of multi-period corporate survival probabilities conditional on macroeconomic and firm-specific covariates. We then provide an empirical implementation of the proposed method for about 300 Iran-listed Industrial firms. Our method combines traditional duration analysis of the dependence of default intensity on time varying covariates with time-series analysis of covariates. 3 - Managing Capacity with Optimal Buffer Size Selection Melda Ormeci Matoglu, University of New Hampshire, 10 Garrison Ave., Durham, NH, 03824, United States of America, melda.ormecimatoglu@unh.edu We model the problem of managing capacity and determining optimal buffer size in a BTO environment as a Brownian drift control problem. We seek a policy that minimizes long-term average cost. The controller can, at some cost, shift the processing rate among 2 rates and has the option of rejecting orders and idling. We show that the optimal policy follows a simple policy and determine the optimal policy parameters. We also calculate important policy performance metrics. 4 - Modeling and Predicting Purchasing Behavior with an Erlang-2 Poisson Lognormal Model Giang Trinh, Senior Research Associate, Ehrenberg-Bass Institute, University of South Australia, 70 North Terrace, Adelaide, SA, Australia, giang.trinh@marketingscience.info We note some practical and theoretical shortcomings of the Erlang-2 Poisson gamma mixture model or the condensed NBD, which has been successfully employed for modeling and predicting consumer purchases. We develop a new model, the Erlang-2 Poisson lognormal mixture model, which has a sounder theoretical base. We derive the conditional expectation of the new model and use it to predict future purchases. We show that the new model predicts future purchases better than the condensed NBD model.

157