2015 Informs Annual Meeting

TC24

INFORMS Philadelphia – 2015

4 - Optimal Control of an Inventory System with Stochastic and Independent Leadtimes Mohsen Elhafsi, Professor, University of California, School of Business Administration, 900 University Avenue, Riverside, CA, 92521, United States of America, mohsen.elhafsi@ucr.edu, Saif Benjaafar, Rui Chen We study a continuous review inventory system with stochastic independent leadtimes. Because orders may not be delivered in the same sequence in which they have been placed, characterizing the optimal policy is difficult and much of the available literature assumes a fixed base-stock policy which we show is sub- optimal and can perform poorly. Instead, the optimal policy is state-dependent and specified in terms of an inventory-dependent threshold function characterized by at most m parameters. TC24 24-Room 401, Marriott Search Across Disciplines: Artificial Intelligence and Chair: Nathan Sturtevant, University of Denver, 2280 S. Vine St., Denver, CO, 80210, United States of America, sturtevant@cs.du.edu 1 - The Cyclic Best-first Search Strategy for Branch-and-bound Algorithms Jason Sauppe, University of Illinois at Urbana-Champaign, 201 North Goodwin Avenue, Urbana, IL, 61801, United States of America, sauppejj@gmail.com, Edward Sewell, Sheldon Jacobson, David Morrison The Cyclic Best-First Search (CBFS) strategy is a generalization of best-first search that splits unexplored subproblems over a collection of heaps, referred to as contours. During the search process, CBFS repeatedly cycles through a list of non- empty contours, selecting one subproblem to explore from each during every pass. Contours can be defined in various ways to influence the search process. This talk will present some properties of CBFS and computational results for a variety of problems. 2 - Adding Random Exploration to Search Algorithms Rick Valenzano, Alberta Innovates Centre for Machine Learning, 2-21 Athabasca Hall, University of Alberta, Edmonton, AB, T5K1X4, Canada, valenzan@ualberta.ca In this work, we will use a simple technique called epsilon-greedy node selection to demonstrate the value of enhancing search algorithms with random exploration. Through empirical evaluation, this technique is shown to substantially improve the performance of search-based automated planners. We also formally analyze this technique to demonstrate that algorithms that employ random exploration are more robust to heuristic error. 3 - Exploiting Large Admissible Heuristics in Search Nathan Sturtevant, University of Denver, 2280 S. Vine St., Denver, CO, 80210, United States of America, sturtevant@cs.du.edu Admissible heuristics, which do not overestimate the cost to the goal, are particularly useful in shortest-path search problems, as they can guide the search. They also can, if necessary, guarantee optimal solutions. This talk looks at recent heuristics that are larger than the memory, and suggests ways of exploiting the problem structure to reduce the memory overhead of storing the heuristic. Two methods, bloom filters and value range compression are discussed, along with their tradeoffs. TC25 25-Room 402, Marriott Online Crowds: Crowdfunding and Social Media Sponsor: Information Systems Sponsored Session Chair: Qiang Gao, University of Arizona, 3700 N 1st Ave. #1020, Tucson, AZ, 85719, United States of America, qiangg@email.arizona.edu 1 - A Distant Supervision Approach for Social Media Pharmacovigilance Xiao Liu, University of Arizona, 1130 E. Helen St., Room 430, Tucson, AZ, 85721, United States of America, xiaoliu@email.arizona.edu, Hsinchun Chen Phamacovigilance refers to the science relating to the detection, assessment, understanding, and prevention of adverse drug events. Prior studies showed Operations Research Sponsor: Artificial Intelligence Sponsored Session

social media can be used to identify adverse drug events with supervised learning approaches. However, they require expert annotation and are not scalable for large datasets. In this study, we develop a framework for pharmacovigilance in social media using distant supervision. Our framework achieves competent performance without annotation. 2 - The Effect of Rating System Design on Negativity Bias Ying Liu, Arizona State University, 1201 S McClintock Dr, Apt 221, Tempe, AZ, 85281, United States of America, yingliu_is@asu.edu, Pei-yu Chen, Kevin Hong Does rating system design affect consumers’ negativity bias in reporting product ratings? We examine the effect with both observational and experimental study. Results suggest that consumers tend to reflect their experiences in the least satisfied dimension in single-dimensional rating systems, whereas the overall ratings in multi-dimensional systems tend to reflect consumers’ average experience. The study suggests that multidimensional rating systems could mitigate consumers’ negativity bias. 3 - Predict Campaign Quality: An Empirical Analysis of the Value of Video in Crowdfunding Markets Qiang Gao, University of Arizona, 3700 N 1st Ave. #1020, Tucson, AZ, 85719, United States of America, qiangg@email.arizona.edu, Mingfeng Lin Videos are prevalent in crowdfunding campaigns where there is usually little verifiable information. Yet to date there is virtually no systematic study of its roles in this new context. We investigate how video features predict campaign quality using data from a leading rewards-based crowdfunding website by implementing both explanatory and predictive models. 4 - Content Monetization in Social Media: Estimation of Demand and Supply for User Generated Content Social networking is reaching a maturity stage with fewer new registrations but more user churning. Our study investigates how a new market mechanism, content monetization, reduces turnover rate by using data from the largest Chinese social network Sina Weibo. It examines the factors that affect both the demand and supply for user-generated content (UGC) in social media. Our results confirm that this nascent mechanism effectively motivates the supply for UGC and also improves its quality. TC26 26-Room 403, Marriott Gray Market, Sustainability, Competition, and Diffusion Cluster: Operations/Marketing Interface Invited Session Chair: Samar Mukhopadhyay, Professor, Sungkyunkwan University- GSB, 25-2 Sungkyunkwan-ro, Jongno gu, Seoul, 110 745, Korea, Republic of, samar@uwm.edu 1 - Countering Gray Market Threat using Marketing Effort Samar Mukhopadhyay, Professor, Sungkyunkwan University- GSB, 25-2 Sungkyunkwan-ro, Jongno gu, Seoul, 110 745, Korea, Republic of, samar@uwm.edu, Xuemei Su Gray markets are likely when there is a significant price difference of the same product in different markets. This paper studies the role of an important variable, marketing effort, in fighting gray market, in addition to price. We find that when both marketing effort and prices are controlled, the manufacturer’s profit is improved. Sometimes, it may even be better not to sell through the authorized channel, but to manage the gray market by controlling the marketing effort levels and prices. 2 - Designing Sustainable Products under Co-Production Technology Yen-Ting Lin, University of San Diego, School of Business Administration, 5998 Alcala Park, San Diego, CA, United States of America, linyt@sandiego.edu, Haoying Sun, Shouqiang Wang We consider a manufacturer who takes a natural resource to make two products through co-production technology. Some consumers are green and additionally value conservation of the natural resource. We show that increasing the portion of green consumers may actually elevate resource consumption. Ruibin Geng, Zhejiang University, 388 Yuhangtang Road, Hangzhou, ZJ, 310058, China, grace.bin1207@gmail.com, Bin Zhang, Paulo Goes

326