Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WC26

4 - Data-driven Quickest Detection of Customer Churn Roozbeh Yousefi, Queen’s University, Kingston, ON, K7M9H1, Canada, Jue Wang In many service industries, customers may stop returning for service without informing the service provider. It is important for the service provider to detect such soft churn as quickly as possible, so that appropriate retention can be made. We develop a churn detection model based on partially observable Markov decision processes in which both transition and observation probabilities are unknown. The optimal policy must balance churn detection with parameter learning. We characterize the structure of the optimal policy and show that the infinite-dimensional belief space can be collapsed to two-dimension, making the optimal policy computationally feasible. 5 - Sales & Operations Planning Linear Programming Model Used by Semiconductor Manufactures Frank Muldoon, Operations Research Application Engineer, Applied Materials, 5225 W. Wiley Post Way, Suite 275, Salt Lake City, UT, 84116, United States In today’s competitive environment, semiconductor manufacturers face the unending challenge of planning for ever-changing customer demands. Factory managers must plan for new orders within existing demands in order to avoid capacity shortfalls of available resources. We propose a linear programming formulation that models this capacity planning challenge in Applied Material’s Advanced Productivity Family software. The result of this optimization identifies demands that cannot be satisfied and resource capacity deficits week, months, or even years in advance giving planners time to adjust. We discuss the implementation of our system within a semi-conductor assembly and test facility. n WC26 North Bldg 132A System Design for Emergency Medical Services Sponsored: Service Science Sponsored Session Chair: Ozgur M. Araz, University of Nebraska, Lincoln, NE, 68588- 0491, United States 1 - Emergency Service Network Design with Constraint Relaxation Adrian Ramirez Nafarrate, Professor, Universidad Panamericana, Zapopan, Mexico, Ozgur M. Araz, John W. Fowler Some emergency scenarios require deploying a limited amount of resources to deliver service within a target time. However, the available resources in large- scale scenarios may make the location-allocation problem infeasible. We propose a modeling approach that represents this situation and present a set of algorithms to relax capacity at each service site and/or time to complete the service operations. The results show how decision makers can incentivize using one type of resource and when is better to balance adding more capacity or service time. In addition, we analyze how the resource allocation strategy interacts with the number of open sites. 2 - Bed Forecasting System in the Emergency Healthcare Network Shao-Jen Weng, Tunghai University, Taichung City, Taiwan Because of high complexity in Emergency Departments, the ability to analyze the present and future status of resource availability is difficult. The goal of this research is to use simulation to create a cloud-based forecasting system called an available bed forecasting system (ABFS) in a regional hospital network in Taiwan in order to provide near-contemporaneous information to Emergency Department personnel and to EMTs in an effort to efficiently predict the optimal hospital to which patients should be transported based on bed and resource availability. 3 - Forecasting Influenza for Emergency Planning Zeynep Ertem, University of Texas at Austin, Austin, TX, 78723, United States Timely and accurate estimates of influenza prevalence, particularly of severe cases requiring hospitalization, can improve control measures to deal with medical emergencies. Our framework uses multi-linear regression to combine forecasts from multiple data sources and optimization with forward selection to choose the most predictive combinations of data sources. We show that the systematic integration of complementary data sources can improve forecast accuracy over single data sources. The optimal combination of predictors includes public health surveillance data and commercially available electronic medical records, but neither search engine nor social media data.

n WC27 North Bldg 132B Information Systems I Contributed Session Chair: Fengmei Gong, University of La Verne, La Verne, CA, 91750, United States 1 - Do Organizational Controls Influence the Effectiveness of Waste Management Green IT Systems? Xue Ning, University of Colorado Denver, 1475 Lawrence Street, Denver, CO, 80202, United States, Dobin Yim, Jiban Khuntia This study explores how organizational controls influence the effectiveness of life cycle assessment systems (LCAS) for waste management in hotels. Effectiveness is measured through total waste generated by hotels and waste management cost reduction. Analysis of secondary data for 1,277 hotels of a large multinational hotel chain is used to find evidence for hypotheses. Study informs to green Joshua Becker, PhD Candidate, University of Pennsylvania, 3620 Walnut Street, Suite 200, Philadelphia, PA, 19104, United States, Damon Centola Groups can form remarkably accurate beliefs even when group members are wildly inaccurate, an empirical phenomenon known as the “wisdom of crowds. This accuracy gain, however, offers no direct benefit to individual group members. Moreover, theoretical accounts argue that individuals must be independent to preserve group accuracy, preventing social learning. We test an alternative theory predicting that structured information exchange networks allow social learning while preserving group accuracy. Experimental subjects made financial forecasts before and after learning the beliefs of peers. Information exchange produced a 25% decrease in error without harming the wisdom of the crowd. 3 - Timely Production of Crowdsourced Answers in Stack Overflow Orcun Temizkan, Ozyegin University, Cekmekoy Kampusu, AB2 324, Nisantepe Mah. Orman Sok. No: 34-36, Istanbul, 34794, Turkey, Ram Kumar Question answering websites that involve volunteer participants such as Stack Overflow are becoming increasingly popular. They represent a crowdsourced knowledge creation process, and thus are large repositories of knowledge. In addition to the quality of answers, a process that results in high quality answers in a relatively short time frame is also important. Therefore, we focus on the generation of high quality answers in a timely manner. We present a survival analysis model of generating high quality answers. We also discuss research and managerial implications. 4 - Are Online Ratings Relative in Competitive Markets? A Mean-field Estimation Approach Mingwen Yang, University of Texas at Dallas, Richardson, TX, 75080, United States, Vijay S. Mookerjee, Eric Zheng, Hongyu Chen We study user ratings in China’s tourism market, where different tours compete with one another for business. The focus is on how user ratings affect the sales of competing tours and how the ratings of these tours are optimally managed over time. We estimate the joint evolution of tour ratings and sales assuming that tour managers act rationally to maximize profit over time. The data for the study comes from sales and review information collected over time for tours listed on Ctrip.com, the largest travel aggregator in China. Our estimation approach uses controlled diffusion processes to jointly estimate two stochastic processes for each tour: sales and review ratings. 5 - The Limitation of Current Digital Gamification Systems and the Alternative Strategy for Sustainable User Engagement Jaewan Lim, PhD Student, University of Maryland, Baltimore County, 1000 Hilltop Cir, Baltimore, MD, 21250, United States In a large number of studies, it has been demonstrated that digital gamification strategy improves the engagement of customer. However, some gamification systems expose short life cycle pattern, resulting in the decrease of the system effectiveness. There is a paucity of literature on this problem. In this regard, the focus of this research is to identify the limitation of current digital gamification systems and propose the model for increasing the retention rate and sustainable user engagement. 6 - The Impact of Information Technology on Technical EfficiencyùEvidence from Industry Fengmei Gong, University of La Verne, 1950 Third Street, La Verne, CA, 91750, United States, Tong Zeng, Yingxia Cao This study intends to examine the impact of Information Technology (IT) investment on technical efficiency across different industries in the U.S. economy. Previous related studies are mostly based on firm level and country-level evidence, this study adds to the literature by providing industry-level evidence. information technology and business value streams of research. 2 - Can Individuals Benefit from the Wisdom of Crowds?

478