Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WB62

2 - A Bi Objective Production Planning and Scheduling Model Dealing with Reworking the Perishable Items in a Parallel Machine System Elham Taghizadeh, PhD candidate, Wayne State University, Detriot, MI, 48202, United States, Farshid EvazAbadian, Setareh Torabzadeh, Abdollah Mohammadi Production systems could always struggle with defective products which can be due to human error, machine break-down, and imperfect production system. Rework is one of the strategies to deal with this issue; rework planning involves the production planning and scheduling of the defected items along the master production planning. This paper presents a multi-objective mathematical model to determine how the work and rework must be scheduled in a parallel machine system for a perishable product. The model objective functions are the minimization of cost and makespan. A linear model is developed and numerical experiments are done based on a case study from the literature, which concerns a biopharmaceutical industry. To find out the importance of each parameter, a sensitivity analysis of the results on varying model parameters is presented as well. 3 - Dynamic Capacity Planning and Matching: A Distributionally Robust Approach Zhaowei Hao, National University of Singapore, Singapore, 119245, Singapore, Long He, Zhenyu Hu, Jun Jiang In this paper, we develop a distributionally robust model to study how a firm makes initial capacity and allocation decisions with upgrading in a multi-product multi-period setting. Using the enhanced linear decision rule (ELDR), we derive a tractable approximation, and develop a theoretical performance guarantee in solving the robust model. Finally, our numerical study shows the performance and computational efficiency of the ELDR solution in benchmark with the solution from dynamic programming. n WB62 West Bldg 103A Joint Session DM/Practice Curated: Data Science in Health Care II Sponsored: Data Mining Sponsored Session Chair: Mehrad Bastani, Stanford University, Mountain View, CA, 94040, United States 1 - Predicting Post Operative Survivability of Lung Cancer Patients a Machine Learning Approach Alireza Fazli Khalaf, Student, Binghamton University, Vestal, NY, 13905, United States, Yong Wang Thoracic surgery is one of the treatment options for lung cancer. Risk measurement of such surgery is critical to choose the operation or not. In this paper, one-year survivability of lung cancer patients after a thoracic surgery is investigated using machine learning techniques. As the study deals with imbalance dataset, synthetic minority over-sampling technique is used to resolve the problem. Next, the performance of five classification methods, namely na ve Bayes, radial basis function network, support vector machine, logistic regression and random forest, on predicting the survivability are compared. Results show that Random forest outperforms other classifiers. 2 - Automated Surgical Term Clustering Used in Unstructured Surgery Descriptions Tannaz Khaleghi, Graduate Teaching Assistant, Wayne State University, 4815 4th Street, Suite 1067, Detroit, MI, 48201, United States, Alper E. Murat Text mining tools provide us a unique opportunity to extract information from textual data. The information therefore can be useful when predicting procedure code and surgery room duration for different surgical cases as text can provide important details about the procedure while other common features might not target them. In this study the goal is to find most informative text feature from unstructured principal procedure and some physician notes by heuristic integrated text mining method which best organize medical text feature set and sets up feature dimensionality reduction efficiently. The output improves surgery code prediction accuracy and produce more reliable surgery duration.

n WB60 West Bldg 102B Dynamic Resource Allocation Problems in Healthcare Sponsored: Health Applications Sponsored Session Chair: Shan Liu, University of Washington, Seattle, WA, 98195-2650, United States 1 - Resource Allocation for Hepatitis C Elimination Qiushi Chen, PhD, Penn State, University Park, PA, United States, Turgay Ayer, Jagpreet Chhatwal More than 70 million people are chronically infected with hepatitis C virus (HCV) globally. With the recent availability of new treatments, the World Health Organization (WHO) set an ambitious target to eliminate HCV by 2030. However, high treatment cost and unawareness of infection remain major barriers to elimination. In this study, we develop an HCV transmission model that aids in optimal allocation of resources to scale-up HCV screening and treatment that can lead to HCV elimination. We present the optimal allocation policies in different health care settings and target population profiles. 2 - Optimal Control of an Infectious Disease with Drug Resistance Naveed Chehrazi, McCombs School of Business, 2110 Speedway, Stop B6000, Austin, TX, 78712, United States, Lauren Cipriano, Eva A. Enns We study the optimal treatment policy for an SIS-type infectious disease with drug resistance. We prove that the optimal policy is bang-bang with a single switching time and the optimal value function may not be Lipschitz continuous. Using numerical analysis, we demonstrate that the optimal policy changes form when the disease transmission rate is a function of disease prevalence (e.g., as a result of social distancing). Our solution approach can be generalized to other control problems. 3 - Prevention of Seasonal Influenza Outbreak via Healthcare Insurance TingYu Ho, University of Washington, Seattle, WA, 98125, United States, Zelda B. Zabinsky, Paul A. Fishman, Shan Liu To prevent the outbreak of seasonal influenza, we develop an integrated insurance mechanism, including vaccination incentives and cost-sharing policies, and formulate the dynamic interaction between a single insurer and multiple insureds as a Stackelberg vaccination game. We implement agent-based simulation modeling with active learning and simulation optimization to determine the interventions that control the spread of flu. Results indicate that incentives and cost-sharing can effectively improve vaccination uptake and maintain low incident rate even with a highly contagious flu. 4 - Allocating Resources for Outreach Programs in Infectious Disease Systems with Information Propagation Sze-chuan Suen, University of Southern California, Los Angeles, CA, 90089-0193, United States, Brian Wilder, Milind Tambe Outreach programs play an integral role in infectious disease control by increasing disease awareness, screening rates, and treatment uptake. Outreach campaigns may reply on information propagation through word of mouth and social media (viral marketing), which can exponentially expand the reach of the original outreach message to preferentially reach targeted groups. We study the optimal allocation of limited outreach resources in the context of tuberculosis in the context of a heterogeneous population with such information spread. n WB61 West Bldg 102C Supply Chain Optimization I Contributed Session Chair: Zhaowei Hao, National University of Singapore, Singapore 119245, Singapore 1 - Coherence of Demand Forecasting: Application to Tompkins International Company Sajjad Taghiyeh, North Carolina State University, Raleigh, NC, 27606, United States, David Lengacher, Robert Handfield We intend to develop a robust multi-tier multi-phase forecasting model for a third-party logistics company. Demands can be represented as the volume of individual Stock Keeping Units (SKUs), and the total volume of SKUs in each distribution center (DC). We aim to use these forecasts and find the aggregation of them to provide a more accurate demand forecasting framework.

463