2016 INFORMS Annual Meeting Program

TD73

INFORMS Nashville – 2016

TD73 Legends A- Omni Operations Management IV Contributed Session Chair: Jose M. Merigo, University of Chile, Av. Diagonal Paraguay 257, Santiago, 8330015, Chile, jmerigo@fen.uchile.cl 1 - Understanding And Managing Sequences Of Alignment Between Technologies And Adopters: Case Research Of Implementations Of A Health Screening Program Jose Coelho Rodrigues, Researcher, INESC TEC and Faculty of Engineering, University of Porto, Rua Dr Roberto Frias, Porto, 4200, Portugal, jose.c.rodrigues@inesctec.pt, Ana C Barros, João Claro Misalignments (lack of compatibility) between technologies and adopters cause productivity losses in early stages of implementation projects. Alignment management is particularly challenging when the adopter is a network of organizations. We use multiple case research of implementations of a health screening program in networks to understand how alignment efforts are sequenced, focusing on the non-linear and cascading sequences. We provide guidelines to improve implementations’ performance, by addressing why non- linear and cascading sequences occur and what are their impacts on such projects. 2 - Mapping Production And Operations Management With VOS Viewer Jose M. Merigo, University of Chile, Av. Diagonal Paraguay 257, Santiago, 8330015, Chile, jmerigo@fen.uchile.cl, Claudio Muller, Sigifredo Laengle The VOS viewer is a computer software that visualizes the bibliographic material through different bibliometric indicators. This study develops a visualization of production and operations management research by using the VOS viewer. The analysis considers bibliographic coupling, co-citation, co-occurrence of keywords and co-authorship for journals, documents, authors, institutions and countries. The results indicate that this field is very diverse with two main cores focused on engineering and management. Researchers from all over the World are making important contributions in the field although the USA is still the leader. Chair: Xiang Gao, University of Minnesota, 111 Church Street SE, Minneapolis, MN, 55455, United States, gaoxx460@umn.edu 1 - Auction Algorithms For Distributed Integer Programming Problems With A Coupling Cardinality Constraint Ezgi Karabulut, Georgia Institute of Technology, 755 Ferst Drive, NW, Atlanta, GA, 30308, United States, ezgi.karabulut@gatech.edu, Shabbir Ahmed, George L Nemhauser We are interested in optimizing discrete problems that use a common resource, namely integer programming problems coupled with a cardinality constraint. Our auction algorithm finds the optimal resource allocations when individual problems are concave. When the problems are not concave, but rather have a concave approximation; and we provide respective error bounds for the auction algorithm. 2 - Fuzzification Of Search Techniques For Linear And Nonlinear Optimization Paul Eugene Coffman, Technical Leader, Virtual Manufacturing and O.R., Ford Motor Company, 6100 Mercury Drive, Dearborn, MI, 48126, United States, gcoffman@ford.com, Stephany Coffman-Wolph Using a three-step framework any algorithm can be converted into an equivalent abstract version known as a fuzzy algorithm. This goes beyond simply converting the raw data into fuzzy data by converting both operators and concepts into their abstract equivalents. Although precision may be reduced, it can be counteracted by gains in computational efficiency. This presentation will discuss linear and non-linear search algorithms that can benefit from fuzzification, results within the context of potential applications, and the characteristics of an algorithm where fuzzification can be utilized. TD74 Legends B- Omni Optimization Methodology IV Contributed Session

3 - Non-stationary Regret Analysis For A Non-convex Online Learning Model Xiang Gao, University of Minnesota, 111 Church Street SE, Minneapolis, MN, 55455, United States, gaoxx460@umn.edu, Xiaobo Li, Shuzhong Zhang In this talk we present a non-stationary regret analysis for an online learning model with smooth but non-convex cost functions. The cost functions are assumed to satisfy a condition which is more relaxed than the usual pseudo- convexity. Moreover, the cost functions are assumed to satisfy an error bound condition, which is implied by the analyticity. Under this framework, assuming only the loss function values can be evaluated we design a learning algorithm without the gradient information, and show that the regret of the algorithm is proportional to the square root of the product of learning periods and the variational budget which is the total variation of the optimal solutions measured in distance.

TD75 Legends C- Omni Behavioral Operations IV Contributed Session

Chair: Junlin Chen, Associate Professor, Central University of Finance and Economics, 39 South College Road, Haidian District, Beijing, 100081, China, chenjunlin@cufe.edu.cn 1 - Manufacturer Salespersons Relationships In Global Markets Considering Inventory Policies And Cultural Effects Sepideh Alavi, PhD Candidate, University of Wisconsin Milwaukee, 1559, N Prospect Ave. Apt 309, Milwaukee, WI, 53202, United States, alavi@uwm.edu The influence of salespersons’ intermediary behaviors on customer retention has encouraged the manufacturers to develop and monitor strategies to increase loyalty in salespersons (Keiko Yamakawa, 2002).Also, cultural types reflect different trust characteristics in their relationship. Little is known about the impacts of culture in manufacturer- salespersons’ relationships. This paper intends to address this gap by investigating the research questions: What are the inventory- related policy factors that enhance manufacturer-salespersons’ relationship? And does culture play a role in the manufacturer- salespersons’ relationship? 2 - Prediction Of SNS User’s Behavior Preference Peng Zhu, Nanjing University of Science and Technology, School of Economics and Management, 200 Xiaolingwei Street, Nanjing, 210094, China, p.zhu@outlook.com Analysis and prediction of user behavior has become significant means to enhance the user experience in Social Networks Services(SNS). However, due to features of social networks, the limitations of user’s time and energy, the social relationships of most social users are incomplete and sparse, it restricts the coverage and accuracy of user behavioral prediction. In response to these problems, this paper extracts user potential social relationship, and by making use of user preference information, it designs effective user preference consistency algorithm. Meanwhile, it proposes a visualizer evaluation method, which also can evaluate the performance of prediction algorithm from micro level. 3 - Strategic Consumer Behavior In Single Rider Lanes At Adventure Parks Arpit Goel, PhD Student, Stanford, 475 Via Ortega, Huang Engineering Center, Stanford, CA, 94305, United States, argoel@stanford.edu Adventure park rides often have separate lanes for single riders. Single riders are added to any ride tram which is not fully occupied, which increases the efficiency of the queuing process. Thus single rider lanes are usually served much faster than regular lanes. But often these lanes are strategically used by families to expedite their waiting times, the risk being the family not being able to take the same ride tram. We model this scenario as a stochastic process, understanding the strategic tradeoffs, showing situations where this strategic behavior significantly harms the welfare, and thereby implying some managerial ideas for adventure parks to further improve their queuing process. 4 - Crowding-out And Overjustification Effects On Pro-social Behaviors: A Quasi-experimental Study Dandan Qiao, Tsinghua University, HaiDian District, Beijing, China, qiaodd.12@sem.tsinghua.edu.cn, Shun-Yang Lee, Andrew B Whinston, Qiang Wei We explore how external incentives would influence one’s pro-social behavior both in the short term and in the long run. Using a large data set on Amazon product reviews (1997-2014), we design a quasi-experimental approach by combining a propensity score matching (PSM) and a difference-in-differences (DiD) method. Several novel measures are proposed to capture reviewers’ writing style and quality by applying linguistic, language processing, and machine learning techniques. Through estimating a series of fixed-effect DiD models, we find evidence consistent with reciprocity, crowding-out, and overjustification effects.

358