Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WC67

2 - Optimal Return Policy in the Presence of Social Networks Ehsan Salimi, University of Florida, 376 Weil hall, Gainesville, FL, 32603, United States, Sina Ansari When a firm allows the return of purchased item, it provides customers with the option of keeping or returning the item. While the option to return item leads to an increase in gross revenue, it may also create additional costs including the social costs associated with returns. In this paper, we study the optimal return policy considering customers’ returns behavior in the presence of social networks. Our findings have important implications for the coordination of marketing and operations decisions. 3 - An Anatomy-adjusted Quality Control Tool for Cancer Radiotherapy Plan Evaluation Arkajyoti Roy, The University of Texas at San Antonio, San Antonio, TX, United States Arkajyoti Roy, Bowling Green State University, Bowling Green, OH, United States, Dan Cutright, Mahesh Gopalakrishnan, Arthur Yeh, Bharat B. Mittal A quality control tool is proposed that allows clinicians to evaluate and directly compare cancer radiation treatment quality of a large set of patients, after accounting for variations in patients’ anatomies. The effect of the inter-patient variations is accounted for through the use of anatomy-adjusted I-Charts. 69 head-and-neck cancer cases are used for the evaluation of the proposed tool. 4 - Statistical Monitoring of Inhomogeneous Continuous Time Markov Chains Yanqing Kuang, University of South Florida, Tampa, FL, United States, Devashish Das, Jianguo Wu, Mustafa Y. Sir In this presentation, we propose a nonparametric scheme for monitoring inhomogenous continuous time Markov chains with a large state space. The proposed framework is used to monitor the timeliness of the healthcare delivery process using time-stamped clinical event sequences. n WC64 West Bldg 104A Joint Session DM/Practice Curated: Data Science for Transportation Related Applications Sponsored: Data Mining Sponsored Session Chair: Sang Min Lee, Korea University, Seoul, Korea, Republic of 1 - Detecting Crash Severity in Passenger Vehicle a Machine Learning Study Rupesh Agrawal, Research Assistant, Oklahoma State University, 700 N. Greenwood Ave, Tulsa, OK, 74106, United States, Robert Fritts, Dursun Delen In 2016, National Highway Traffic Safety Administration reported nearly a trillion-dollar impact from the loss of productivity, loss of life, and other consequences related to automobile crashes. This study seeks to enhance the current body of knowledge in discovering variables impacting the level of injury severity in passenger vehicle accidents using variable selection and data balancing techniques (i.e., oversampling and undersampling) using multidimensional, feature-rich, and highly-structured data with Machine Learning algorithms. 2 - Conditional Monitoring of Wheel Wears for High-speed Trains: A Data-driven Approach Peiwen Xu, PHD Candidate, City University of Hong Kong, 88 Tat Chee Avennue, AC1, Hong Kong, Hong Kong, Weiran Yao, Yang Zhao, CAI YI, Lishuai Li, Jianhui Lin, Kwok-Leung Tsui The rapid expansion of high-speed railway network is placing increased emphasis on the optimization of the maintenance process to enhance the availability and efficiency of the train system with a high standard of safety and reliability. A data- driven method is proposed to monitor wheel wears in a high-speed railway system. The result can provide an early warning of a component degradation, enabling the switch from fixed interval maintenance to condition-based maintenance. The proposed method combines signal processing and statistical methods to extract relevant information from vibration data and then predict wheel wears. The accuracy of this method is tested by real operational data in China. 3 - Incremental Learning for Nonstationary Traffic Control in Automated Vehicle Systems Sang Min Lee, Korea University, 221 New Engineering Building, Korea Univ., 145 Anam-ro, Seongbuk-gu, Seoul, Korea, Republic of, Sung Ho Park, Seoung Bum Kim We introduce an incremental learning method for adaptive traffic control in a large-scale automated vehicle system. We present a change-aware learning method that combines a change detector with adaptation algorithms. To demonstrate the effectiveness of the proposed method, we conducted an experimental study to evaluate the predictive performance using the high-fidelity simulator.

n WC66 West Bldg 105A Reliability I Contributed Session

Chair: Yuan Chen, Ohio University, Athens, OH, 45701, United States 1 - Assessing a Maintenance Outsourcing Production System under Presence of Cyber Threats and Channel Coordination Anh Ta, University of North Texas, Denton, TX, 76205, United States, Hakan Tarakci, Shailesh S. Kulkarni, Victor R. Prybutok Outsourcing of maintenance is desirable in many production systems. The Internet of Things (IoT) enabled “smart manufacturing systems but the resulting connected format makes these systems potential targets for cyber-attacks. This study assesses maintenance outsourcing for potential cyber threats as well as channel coordination cost subsidization. We provide guidelines and quantify the effect of various monetary, reliability and cyber-attack parameters. 2 - Production Control to Reduce Waste Productions in a Two-machine-one-buffer Bernoulli Serial Line Penghao Cui, Northwestern Polytechnical University, NO.127, West Youyi Road, Xi’an, XiÆan, 710072, China, Junqiang Wang We study production control problems in a two-machine-one-buffer Bernoulli serial line including waste production. In such a system, the second machine produces a certain number of non-standard quality productions each time it restarts after a stop either a failure or a starvation. Using a Markov model, an optimal control policy is developed to restore the buffer to a predefined threshold each time it gets empty for reducing starvation frequency of the second machine. Closed-form expressions of performance measures are derived and the effect of the policy is carried out by comparing with a serial line without it. 3 - Reliability-redundancy Optimization for Continuous-state Series-parallel System with Degrading Components Yuan Chen, Ohio University, Athens, OH, 45701, United States, Tao Yuan This presentation discusses a reliability-redundancy optimization problem for continuous-state series-parallel systems consisting of degrading components. A structure function is introduced to describe the relationship between the state of the system and the states of the components over time. An optimization model that maximizes the reliability of the system subject to a cost constraint is proposed to find the optimal redundancy design. A battery pack system for electric vehicle applications is employed to illustrate the proposed methodology. n WC67 West Bldg 105B Joint Session ISS/SMA: Social Media and Information Systems Sponsored: Information Systems Sponsored Session Chair: Xue Tan, Indiana University, Bloomington, IN, 47405, United States 1 - Can Your Facebook Page Likes Predict Your Dating Behavior? Behnaz Ghahestani Bojd, University of Washington, Seattle, WA, 98105, United States, Yong Tan Many existing dating apps require users to login with their social media accounts. For example, apps like Tinder require a Facebook login to use its service. This grants the apps access to select Facebook data, which speeds up the process of creating dating profiles. It also helps companies detect fake profiles. In this study, using data from an online dating app, we study the relationship between users’ Facebook page likes and their dating preferences, and app usage. Although users do not observe each other’s Facebook page likes, their interests affect their dating judgments. 2 - Linking Clicks to Bricks: Spillover Benefits of Online Advertising Mi Zhou, Carnegie Mellon University, Pittsburgh, PA, 15217, United States, Vibhanshu Abhishek, Edward Kennedy, Kannan Srinivasan, Ritwik Sinha Businesses have widely used email ads to send promotional information to consumers. While email ads serve as a convenient channel to target consumers online, are they effective in increasing offline revenues for firms that predominantly sell in physical stores? Is the effect heterogeneous across different consumer segments? If so, on which consumers is the effect highest? We address these questions using a big dataset from a large domestic retailer. Using a doubly robust estimator that incorporates machine learning methods for causal estimation, we find email ads can increase a consumer’s spending in physical stores by $11.8, and the effect is highest among those with fewer interactions recently.

491