Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WA28

1 - Development of a Context-aware Serendipitous Recommendation System

n WA25 North Bldg 131C Service Queueing Sponsored: Service Science Sponsored Session Chair: Thomas Robbins, East Carolina University, 3212 Bate Building, Greenville, NC, 27858, United States 1 - Evaluating Tail Performance of Queueing Systems Using a Novel Stochastic Ordering Approach Hung T. Do, Assistant Professor, University of Vermont, 55 Colchester Ave., 207 Kalkin Hall, Burlington, VT, 05405, United States, Masha Shunko, Alan Scheller-Wolf In some service systems, performance in the right tail of the queue size and/or waiting time distribution is the main source of managerial concern. We propose a new perspective on analyzing performance of such systems by defining new stochastic orders to compare the tail of distributions and performance measures. 2 - Call Center Staffing with Uncertain Dependent Arrival, Service, and Abandonment Rates Tahir Ekin, Assistant Professor of Quantitative Methods, Texas State University, 601 University Dr. Mccoy 451, San Marcos, TX, 78666, United States, Tevfik Aktekin Many stochastic service models assume the system (arrival, service, and abandonment) rates to be deterministic inputs and those that treat them as uncertain assume they are independent. We propose methods for determining the optimal number of servers of a service system while modeling the system rates as dependent random variables. In doing so, we take the Bayesian point of view of inference and obtain joint posterior distributions of these system rates. Then, we utilize an augmented probability simulation based optimization method to solve the resulting decision model. We illustrate the proposed method and implications of ignoring dependence in system rates using real world call center data. 3 - Balancing Service Speed and Agent Utilization for an Online Insurance Platform Andres I. Musalem, U. of Chile, Beauchef 851, Santiago, 8370456, Chile, Marcelo Olivares, Daniel Yung We study an online platform that sells car insurance policies. Customers who do not purchase online are then contacted via telephone. We find that as the time that it takes the platform to contact a customer increases, the probability that the customer purchases an insurance policy decreases. Hence the platform might gain from improving the speed at which it contacts its website visitors, for example by hiring more agents, which however affects their utilization and compensation. We empirically study this trade-off for the platform where on the one hand a greater call center capacity leads to better speed of contact but also reduces the incentives for experienced agents to keep their jobs at the platform. 4 - An Exact Algorithm for the Service Bundles Design Problem Yifu Li, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, 999077, Hong Kong, Xiangtong Qi The service sector has become to dominate the global economy and evolves at a fast speed. Among the vast category of the service product, experiential service has become an indispensable part of our daily life. A crucial problem is how to create service bundles with given activities in considering both customer experience and practical constraints. Based on the results of the experiential studies, we model and analyses the service bundles design problem, which is proved to be a binary sum of ratios problem. We develop a geometric branch and bound algorithm to solve it to optimality and improve the algorithm with several techniques. It has been tested to be more efficient than the traditional algorithms. 5 - The Impact of Order Ahead Apps on Fast Food Queues Thomas Robbins, Associate Professor, East Carolina University, 3136 Bate Building, Greenville, NC, 27858, United States Order ahead mobile apps are becoming increasingly popular at Quick Serve Restaurants. These apps allow customers to pre-order their meal and skip the line when arriving at the store. In this talk we investigate the hybrid queuing system created by these applications. Using discrete event simulation we examine the impact these apps can have on system performance for customers who use the app, customers who do not, and the system as a whole. n WA26 North Bldg 132A Smart Service Systems with AI and Blockchain Sponsored: Service Science Sponsored Session Chair: Chiehyeon Lim, UNIST, Ulsan, 44919, Korea, Republic of

Changhun Lee, UNIST School of Management Engineering, Ulsan, Korea, Republic of, Gyumin Lee, Chiehyeon Lim Recommendation system development has been an important domain over last several decades. However, as technique has improved and research on recommendation system increased, the importance of developing a “serendipitous” recommendation system has emerged. For that, we paid attention to the latent feature the items are recognized. We assumed that people would move from an item to a next item through a latent feature reflected in a sequence of items. We indirectly show latent features by presenting a topic map, and suggest a context-aware based serendipitous recommendation system. Our interim result seems only context-aware-accurate, but at least showed a possibility of being serendipitous. 2 - Understanding Relations among ERP Success Factors: A Quantitative Approach based on Association Rules Mining Jonghyeon Ko, UNIST School of Management Engineering, Ulsan, Korea, Republic of, Marco Comuzzi The critical success factors (CSFs) of ERP implementation have been widely researched on the definition and the rank of CSFs and categorization according to each type of companies. The CSFs of ERP implementation is related and affected each other so that a lot of experts and researchers have studied different interrelationships among CSFs of ERP implementation and conceptual model of interrelationships. This paper approaches to this problem differently with data- driven model using a company data-set from literatures. The association rule method is used to find interrelationship among groups of CSFs as well as each factor of CSFs. The result shows the entire interrelationships of CSFs with numerical metrics representing the direction and strength of rule. 3 - Mechanism and Development of Blockchain-based Smart Service Systems Chiehyeon Lim, UNIST, Room #606-5, Building #114, Ulsan, 44919, Korea, Republic of, Marco Comuzzi, Byoung Ki Seo In this presentation, first, we define the mechanism how the blockchain technology can contribute to improving service systems. Second, we report R&D project cases on the development of blockchain-based smart service systems. Third, we discuss implications of our work to the Service Science theory and practice. 4 - Smart Services in European Mechanical Engineering Companies Thomas Meiren, Fraunhofer IAO, Nobelstr. 12, Stuttgart, 70569, Germany, Anastasia Tzitamidou Digitally supported services like advanced remote condition monitoring, predictive maintenance, new control and automation solutions as well as profiling and behaviour tracking play an increasing role in mechanical engineering industry. They are making use of the growing volume of data that is being captured every day and are combined in innovative ways in order to create on- demand, personalized solutions for customers. Moreover, product performance and customer behaviours will get visible as they have never been before. The conference presentation will show results from an empirical study within the European mechanical engineering industry. n WA28 North Bldg 221A Practice- Logistics 1 Contributed Session Chair: Josue Velazquez Martinez, Massachusetts Institute of Technology, Cambridge, MA, 02139, United States 1 - A Unified Dynamic Control for Energy-aware Electrical Vehicle Operations Seokgi Lee, Assistant Professor, University of Miami, 1251 Memorial Drive, McArthur Engineering Building, Room 281, Coral Gables, FL, 33146, United States, Hyun Woo Jeon This study investigates the time and cost benefits that can be gained by the integrated control of EMV operations combined with an energy consumption prediction. Dynamic models for EMV charging capacity and operations control systems will be developed and serve as a scientific basis for a unified control algorithm, in which charging capacity, routing, and battery charging schedules of EMV are simultaneously controlled in a real-time manner. An electrical load variation at the warehouse facility level will be predicted by developing a novel electrical load forecasting model based on cutting-edge machine learning techniques, which will serve as important guidelines for unified decision-making.

419