Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WC61

3 - Dialysis Facilities: Does Location, Staffing, and Mode Impact Access to Care? Michael G. Klein, San Jose State University, One Washington Square, San Jose, CA, 95192-0069, United States Kidney failure is treated with dialysis until transplant or death. Regardless of the travel burden, many patients always opt to go to a facility, while some always opt for home dialysis. For others, the choice varies depending on the location of available facilities. Existing patients also switch dialysis modes (e.g. from facility- based dialysis to home dialysis) bringing different requirements for providers over time. I propose a new model to determine the best network of dialysis facilities from an access to care perspective. Through a California case study, I propose to illustrate the model and help identify areas for improvement. 4 - Readmission Reduction Programme for COPD Exacerbation: From Predictive to Prescriptive Analytics Shaochong Lin, City University of Hong Kong, 5022, AC2, Kowloon Tong, Hong Kong, Frank Y. Chen Chronic Obstructive Pulmonary Disease (COPD) is a chronic, non-reversible lung disease. Based on a predictive model, we can early identify high-risk subpopulation with high accuracy for further follow-up interventions. This work aims to use the predictive model to guide allocation of limited health care resources to balance efficiency - readmission reduction and health outcomes. Chair: Anurag Agarwal, University of South Florida, Information Systems and Decision Sciences, Coll of Business, Sarasota, FL, 34243, United States 1 - Optimal Decision of Supply Chain with Consumer Returns in Electronic Commerce Wenjuan Wang, Dongbei University of Finance and Economics, Dalian, China The impacts of the customer’s behaviors can damage the profits of supply chain with customer returns. Some related knowledge of consumer return in supply chain is summarized. We estimate customer demand through a choice model, which incorporates reference price, utility function and customer loss aversion. We analyze the optimal and equilibrium strategies for a seller operating. Our system can improve the efficiency of decision making and provide better customer service for the enterprise’s operation. 2 - Matching and Revenue-risk Sharing Contract with Heterogeneous Risk Averse Supplier and Retailer Hewen Liu, University of Miami, Miami, FL, United States The following analysis covers a decentralized matching model with a bilateral supply chain including risk averse suppliers and retailers. I characterize the stable matchings with endogenous production plan and revenue sharing contract. Depending on the trade-off between the expected revenue and demand uncertainty, both positive assortative matching (PAM) and negative assortative matching (NAM) can occur at a stable matching equilibrium. These results shed light on the role of the endogenous production plan in coordination supply chain allocation and inventory management. 3 - Evaluating the Tradeoff between the Delivery Cost and Customer Lead Time in Last Mile Delivery Sanam Azadiamin, PhD Candidate, Ohio University, Athens, OH, 45701, United States, Dale Masel In supply chain, the purpose of last mile delivery is to deliver items to customers in manageable cost and time. For this purpose, vehicle routing problem has been used to solve the problem considering both delivery cost and customer lead time factors by examining the impact of driver’s batch size on both factors. The travel area’s largeness has also been considered in finding the optimal solution. 4 - Heuristics versus Exact Method Comparison of a Supply Chain Scheduling Problem with Penalties Anurag Agarwal, University of South Florida, Information Systems and Decision Sciences, Coll of Business, Sarasota, FL, 34243, United States, Ramakrishna Govindu Given the complexity of supply chain scheduling problems, they are solved by applying heuristics (using due times and other improvisations) and metaheuristics to find good solutions quickly. Better formulations can help reduce the time for exact solutions. We investigate the performance comparisons between multiple heuristics and an exact method on a supply chain scheduling problem. n WC61 West Bldg 102C Supply Chain Optimization II Contributed Session

n WC62 West Bldg 103A Joint Session DM/Practice Curated: Data Science in Health Care III Sponsored: Data Mining Sponsored Session Chair: Maryam Soltanpour Gharibdousti, Binghamton University, 1120 Murray Hill Road, Vestal, NY, 13850, United States 1 - Studying Impact of Physical Activity on Human Learning Propensity and Emotional Response Modulation using EEG Data Klim Drobnyh, Arizona State University, Tempe, AZ, 85281, United States, Ghazal Shams, Robert Atkinson, George Runger The goal of our study was to investigate the impact of acute and chronic physical activity on learning propensity and emotional response modulation. Two groups of participants with high and low levels of exercise were instrumented with EEGs and tested for cognitive load, and furthermore different stimuli were presented. This generated high-volume, rich datasets, and several analytical tasks were completed to handle this high-volume data. 2 - Hospital Length of Stay Prediction Model for Neurosurgery Inpatients Applying Various Data Mining Techniques Sahar Khamsehi, SUNY Binghamton, 4400 Vestal Pkwy E, Binghamton, NY, 13850, United States Prolonged length of stay (LOS) at hospitals have been controversial topic that leads to extra cost for hospitals. Consequently, it increases patients’ turnaround time. The purpose of this study is to determine non-clinical factors that may prolong length of stay and develop predictor models using contributing factors. The results from three prediction techniques of Artificial Neural Network (ANN), Logistic Regression (LR) and Support Vector Machine (SVM) obtained from neurosurgery database for a period of 2 consecutive years inclusive of 14 non- clinical factors, are statistically compared. Conclusively, logistic regression has the highest accuracy among all data mining techniques. 3 - Survival Rate Prediction in Cardiac Patients with Heart Transplant or Assisted Devices Maryam Soltanpour Gharibdousti, PhD Candidate, Binghamton University, 4400 Vestal Pkwy E, Binghamton, NY 13902, Vestal, NY, 13902, United States, Mohammad Khasawneh The survival rate prediction for the organ transplant surgery patients can help to classify patients risk levels and potential post-surgical complications. The research used the data for cardiac patients with either medical assist devices such as Impella and Left Ventricular Assist Devices (LAVD) or heart transplant patients. The significant factors such as demographic information, baseline patient characteristics, baseline hemodynamics, laboratory values, and in-hospital complications can predict the survival rate after the transplant surgery. The data from one of the Organ Procurement Organizations (OPO) in New York State is analyzed using several machine learning algorithms. n WC63 West Bldg 103B Joint Session DM/Practice Curated: Data Science for Operations and Quality Management Sponsored: Data Mining Sponsored Session Chair: Yanqing Kuang, University of South Florida, Tampa, FL, United States 1 - Data-driven Consumer Debt Collection via Machine Learning and Approximate Dynamic Programming Qingchen Wang, Amsterdam Business School, Amsterdam, Netherlands, Ruben van de Geer, Sandjai Bhulai This paper presents a novel data-driven framework for the optimization of the consumer debt collection process. We consider the problem of scheduling outbound calls made by debt collectors to a portfolio of debtors with heterogeneous and dynamic repayment behavior. We model this problem as a Markov decision process and approximate the value function based on predictions of individual debtors’ repayment probabilities by leveraging historical data and using a state-of-the-art machine learning technique. A controlled field experiment with an industry partner showed an increase in collected cases at a significant decrease in calling efforts compared to their current collection policy.

490