Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MA25

n MA25 North Bldg 131C Machine Learning in Service Sponsored: Service Science Sponsored Session

n MA26 North Bldg 132A CRM Sponsored: Service Science Sponsored Session Chair: Daniel Ringbeck, WHU - Otto Beisheim School of Management, WHU, Vallendar, 1, Georgia 1 - Predictive and Prescriptive Frameworks Leveraging Customer Lifetime Value The success of deterministic customer relationship management tools, such as the recency, frequency, and monetary model, is limited amid randomness in customer transactions and future uncertainties in customer churn. We develop a predictive customer lifetime value model under a competing risk framework that overcomes such limitations. Furthermore, we present a prescriptive resource allocation model for marketing campaigns to target high-valued customers with high-risk of churn. The models are evaluated on a membership-based firm in the hospitality industry. 2 - A Bayesian Network Approach for Predicting Customer Churn Yugang Yu, University of Science & Technology of China, 96 JinZhai Road, Hefei, China, Huan Yu, Libo Li Customer churn prediction is an important business research question for companies to identify potential customer loss. This paper is motivated by a consulting experience with a telecom service provider who try to predict customer churn of different communication operators and channels. The main challenges are the high dimensionality and imbalanced datasets. We develop a novel model which combines a lasso logistic regression for the binary outcome of whether the customers being loyal, with a truncated model of Bayesian network for disloyal customers. Using data from this service provider, our proposed method has a better prediction results compared with other benchmark models. 3 - Organizing Online Crowds for Value Co-creation Zhiyi Wang, National University of Singapore, 15 Computing Drive, COM2-01-02, Singapore, 117418, Singapore, Jungpil Hahn Crowdsourcing has widely been used as a strategy for sourcing ideas and efforts to facilitate innovation and new product development through Information Technology. However, it is challenging for firms to effectively organize online crowds for innovation and value co-creation. In this study, we investigate an online new product invention community and understand what types of crowds are more important to value co-creation process. Our empirical analysis with data mining and econometric modeling identified important types of contributors in the crowd and their influences on innovation outcomes. 4 - Multi-focal Lead Scoring for Predictive Modeling in Business-to-business Market Jingyuan Yang, George Mason University, Fairfax, VA, United States An important aspect of customer acquisition in B2B marketing is lead prospecting which guides the business to select the right leads to pursue. While the current practice often selects leads with ad-hoc basis or pure managerial intuitions, a more systematic and data-driven approach is needed to improve the quality of lead selection by quantifying a specific lead’s tendency to become a customer. To this end, we introduce a multi-focal predictive lead scoring model which can improve the performance of predictive lead scoring. Specifically, leads are first divided into several focal segments. Then, a logistic regression scoring model is learned for each segment with multi-task learning technique. 5 - Proactive Retention Management in Retailing Identifying Predicting and Preventing Partial Defection Daniel Ringbeck, WHU-Otto Beisheim School of Management, Vallendar, Germany, Dmitry Smirnov, Arnd H. Huchzermeier Maintaining strong customer relationships is a top priority for retailers. We develop a framework for proactive retention management based on data from a German hypermarket chain. We propose a novel definition of partial defection in noncontractual settings, demonstrate how to forecast the likelihood of defection of individual customers, and formulate a retention campaign optimization model. Pallav Routh, The University of Texas at San Antonio, San Antonio, TX, United States, Arkajyoti Roy, Jeff Meyer

Chair: Ashish Gupta, IIM Rohtak, MDU Campus, Rohtak, 124001, India 1 - Deep Reinforcement Learning and Transfer Learning for Taxi Driver Dispatching Zhiwei Qin, DiDi Research America, 450 National Ave, Mountain View, CA, 94043, United States, Zhaodong Wang, Xiaocheng Tang, Jieping Ye, Hongtu Zhu Deep reinforcement learning has achieved many successes in solving different types of sequential decision problems. In this work, we propose learning solutions based on deep Q-networks to optimize the dispatching policy for taxi drivers on the DiDi ride-sharing platform. We construct the evaluation environment using real-world spatio-temporal trips data and train dispatching agents for this challenging decision task. Due to problem diversity across different cities, transfer learning is brought in to help increase the learning adaptability and efficiency. We empirically evaluate the performance of our dispatching algorithm and show the benefits of knowledge transfer in the spatial domain. 2 - Modeling Human-robot Collaboration in a Dynamic Workspace as a Markov Decision Process Henry I. Ibekwe, GreenBerry Robotics, 2245 Texas Drive, Suite 300, Sugarland, TX, 77479, United States We present a method for modeling and evaluating the performance of human- robot collaboration for well-defined material handling tasks in a dynamic workspace. Collaborative Robots involves the use of robots to perform tasks alongside humans with the goal of increasing efficiency of the desired task. We model the robot decision-making process as a Markov Decision Process (MDP) where actions executed by the robot are dependent on both the state of the system and the actions performed by the human. An efficient policy for the Markov Decision Process is solved using a dynamic programming algorithm and the policy is implemented for the robot collaborating with a human in a simulated and real environment. 3 - Operational Improvements in Healthcare using Blockchain Sanjeev K. Bordoloi, University of St. Thomas, Opus College of Business, 1000 LaSalle Avenue, TMH 443, Minneapolis, MN, 55403, United States, Sathiyavani Chandran Blockchain is a decentralized protocol that combines transparency, immutability, and consensus properties to enable secure, pseudo-anonymous transactions. Many healthcare organizations are currently experimenting with Blockchain technology, and its use is expected to grow rapidly. We explore how Blockchain can be used in healthcare for more effective operations such as saving time, reducing cost, reducing risk and increasing trust among stakeholders. 4 - Recent Approaches to Creating and Analyzing Big Data in Technology Management and Commercialization Clovia Hamilton, Assistant Professor of Management, Winthrop University, 310 Thurmond Building, Rock Hill, SC, 29733, United States This is a study of three (3) technological advances related to Big Data: web scraping, natural language processing and machine learning. This research identifies ways in which these advances can become more relevant and useful to empirical researchers in the field of technology and innovation management. This research provides a brief look at some applications of these advances in research studies, and engenders a broader discussion around the relevance and impact of these advances to research in technology and innovation management. In particular, making use of these advances in empirical research related to Ashish Gupta, Associate Professor of Analytics, Auburn University, 417 Lowder Hall, 405 West Magnola Ave., Auburn, AL, 36830, United States, Alireza Farnoush, Dolatsara Ahady Hamidreza, David Paradice In this study, we use analytics approaches for understanding the characteristics of companies that have the intention to adopt Blockchain technology. We describe the data acquisition process for identifying companies that have intent to adopt. Subsequently, we use financial indicators of the companies and biographical information of the board members to develop our insights. technology commercialization was researched and is discussed. 5 - Identifying Indicators of Blockchain Adoption

128