2016 INFORMS Annual Meeting Program

MB06

INFORMS Nashville – 2016

MB06 102A-MCC Data Mining in Text Analytics Sponsored: Data Mining Sponsored Session Chair: Onur Seref, Virginia Tech, Pamplin 1007, Blacksburg, VA, 24061, United States, seref@vt.edu 1 - Tracking The Evolution Of User Interests In Online Communities Theodoros Lappas, Stevens University of Technology, tlappas@stevens.edu Online communities are the hubs of our virtual world. A community is typically focused on a broad area, such as sports or politics. Interested users participate in the community by joining discussion threads on relevant topics within the scope of the general theme. In this work, we hypothesize that a user’s level of interest in each topic is correlated with her maturity within the community. We evaluate our hypothesis on datasets from different domains and present a temporal user- interest model. Our study provides insight on the nature of user generated content and has strong implications for any application based on user interests. 2 - A Network-based Model For Conversation Decomposition In Text Mining Sukhwa Hong, Virginia Tech, Blacksburg VA, United States, sukhwa@vt.edu, Onur Seref, Michelle Seref, Alan Abrahams We present a network-based framework to identify and cluster conversational phrases in classes of text data using prevalence scores of n-gram structures and their connections. Our framework extends the “bag-of-words” models by network-based clustering methods to create sub-graphs of connected n-grams. The paths in these sub-graphs represent sequences of words, which form conversational phrases with richer contextual meaning. We use sequence alignment methods to identify variations of these phrases and apply the proposed framework to study a collection of discussion posts from the automotive industry. We compare effectiveness of our method to standard “bag-of-words” models such as LDA. 3 - Two Are Better Than One? An Empirical Study On Crowd Performance For Stock Prediction Hong Hong, Xiamen University, Xiamen, China, hongh@vt.edu Qianzhou Du, Alan G. Wang, Weiguo Fan, Di Xu Online investment communities have been a popular venue for individual investors to share and interact with each other. Prior research confirms the importance of crowd wisdom in the stock markets context, but fails to investigate the impact of crowd characteristics on crowd performance. Guided by the Crowd Wisdom theory, we conduct an empirical study using data collected from Stocktwits to fill this research gap. Our findings show that diversity, independence, and decentralization are positively related to crowd performance. In addition, crowd size significantly moderates the influence of crowd characteristics on crowd performance. This study has both theoretical and practical implications. 4 - An Intelligent Multilayer Hotel Recommender System Ashkan Ebadi, University of Florida, Gainesville, FL, United States, ashkan.ebadi@ufl.edu, Adam Krzyzak Techniques behind the recommender systems have been improved over the time. Recommenders help users to find their required products or services through analysing and aggregating other users’ activities and behaviour. In this paper, we propose an accurate multi-layer hybrid hotel recommender system that uses multi-aspect rating. We used large-scale data of different types and designed a system that is able to suggest hotels which are tailored to the given user. The system employs natural language processing and topic modelling techniques to assess sentiment of users’ reviews. The recommender engine contains several sub- systems where each sub-system contributes to the final recommendations.

3 - A Polyhedral Study Of The Integrated Minimum-up/-down Time And Ramping Polytope Yongpei Guan, University of Florida, Gainesville, FL, United States, guan@ise.ufl.edu, Kai Pan We study the polyhedral structure of an integrated minimum-up/-down time and ramping polytope, which has broad applications in variant industries. By exploring its structures, we derive strong valid inequalities and explore a new proof technique to prove these inequalities are sufficient to describe variant two- period and three-period convex hulls. For multi-period cases, we derive generalized facet-defining strong valid inequalities with efficient polynomial time separation algorithms to improve the computational efficiency. Extensive computational experiments are conducted to verify the effectiveness of our strong valid inequalities. 4 - Integrated Expansion Planning Framework For Interconnected Power Systems; Heat Supply; And Gas Infrastructure Yasaman Mozafari, University of Calgary, Calgary, AB, Canada, y.mozafari@ucalgary.ca, William Rosehart Increasing gas-fired generation capacity and interest in highly efficient combined heat and power generation units (CHPs) in power systems imply interdependencies between electricity, heat, and gas infrastructure. More efficient expansion planning results can be obtained by effectively modeling these couplings in the planning optimization problem. In this work, a comprehensive integrated framework for expansion planning of power systems, heat supply, and gas infrastructure is proposed. Modeling the independencies substantially reduces the cost and GHG emissions incurred in energy sector, which is illustrated through the simulation results for Alberta energy system. MB05 101E-MCC 2016 INFORMS BOM Section Best Working Paper Awards Sponsored: Behavioral Operations Management Sponsored Session Chair: Stephen Leider, University of Michigan, Ann Arbor, MI, leider@umich.edu 1 - A Behavioral Study On Abandonment Decisions In Multi-Stage Projects Javad Nasiry, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, nasiry@ust.hk, Xiaoyang Long, Yaozhong Wu, Yaozhong Wu We experimentally investigate continuation/abandonment decisions in a multi- stage project under two conditions: when the project is reviewed at every stage and when review opportunities are limited. We find systematic deviations from the optimal solution: subjects may wrongly continue or abandon the project, and their decisions are path dependent. We propose a behavioral model which explains the behavioral regularities. 2 - Ideation-Execution Transition In Product Development Evgeny Kagan, University of Michigan, Ann Arbor, MI, kagan@umich.edu, Stephen Leider, William Lovejoy We show experimentally that design performance is significantly worse when designers decide for themselves how to schedule the development process. We demonstrate several remedies for situations when external allocation of time to development phases is not possible. Managers can improve performance by “nudging” individuals towards early physical build, or by requiring them to com- mit to a transition time beforehand. However, the most effective way to improve performance is contingent transition - a requirement to present a prototype that exceeds a minimum performance hurdle. 3 - Impact of Queue Configuration On Service Time: Evidence From A Supermarket Yong-Pin Zhou, University of Washington, Seattle, WA, yongpin@uw.edu, Jingqi Wang We study how queue configuration affects server’s service time using data from a supermarket. We find that servers in dedicated queues are about 10.7% faster than those in pooled queues, after controlling for the queue length, mainly due to a direct social loafing effect. We also demonstrate that pooling has an indirect negative effect on the service time through its impact on the queue length. In aggregation, the social loafing effect dominates and servers slow down (a 6.86% increase in service time) in pooled queues.

149