2015 Informs Annual Meeting

SA25

INFORMS Philadelphia – 2015

SA23 23-Franklin 13, Marriott Queueing Models Sponsor: Applied Probability Sponsored Session Chair: Alexander Stolyar, Lehigh University, 200 W. Packer Ave, Room 484, Bethlehem, United States of America, als714@lehigh.edu 1 - Diffusion Approximations for Large-scale Buffered Systems Tonghoon Suk, Georgia Institute of Technology, 755 Ferst Drive, NW, Atlanta, GA, 30305, United States of America, tonghoon.suk@gmail.com Motivated by a desire to design stochastic systems satisfying certain quality of service, we establish mean-field limit theorems to approximate systems. We start with a simple model consisting of n buffers and a server, and develop diffusion approximations for the system with the randomized LQF scheduling algorithm. We achieve this by allowing the number of sampled buffers d = d(n) to depend on the number of buffers n, which yields an asymptotic decoupling of the queue length processes. 2 - Stability Verification: A Monte-Carlo Approach Neil Walton, University of Amsterdam, Science Park 904, Amsterdam, Netherlands, n.s.walton@uva.nl, Brendan Patch, Michel Mandjes We implement a form of simulated-annealing as a method for detecting if a queueing system is unstable for a given set of loads. 3 - A Large-Scale Service System with Packing Constraints: Greedy-Random Algorithm with Sub-Linear Error Alexander Stolyar, Lehigh University, 200 W. Packer Ave, Room 484, Bethlehem, PA, United States of America, als714@lehigh.edu, Yuan Zhong There are multiple input flows of different customer types and infinite number of servers. Each server may simultaneously serve multiple customers, subject to general “packing constraints”. Each customer leaves after an independent random service time. We consider a version of a Greedy-Random customer assignment (packing) algorithm, and prove that it has an asymptotic competitive ratio 1, as the input flow rates grow to infinity. 4 - Heavy-traffic Behavior of the Maxweight Algorithm in a Switch with Non-uniform Traffic R. Srikant, University of Illinois Urbana-Champaign, CSL 107, Urbana, IL, United States of America, rsrikant@illinois.edu, Siva Theja Maguluri We consider a switch operating under the MaxWeight scheduling algorithm. The traffic pattern is assumed to be such that all ports are saturated, but otherwise can be non-uniform. It is shown that the heavy-traffic scaled queue length under MaxWeight is within a factor of two compared to the smallest achievable queue length.

2 - Mining Hidden Organizational Structures from Meeting Records Jiexun Li, Assistant Professor, Oregon State University, College of Business, Corvallis, OR, 97331, United States of America, jiexun.li@oregonstate.edu, Zhaohui Wu, Bin Zhu Organizations often contain complex structures formed by social relationships. This study introduces an approach to finding hidden structures by mining meeting records. Using text-mining techniques, we extract information about persons and their opinions on topics. We conducted cluster analysis and network analysis to uncover hidden structures within the organization. Our preliminary study shows promising results. We are in the process of improving our approach and conducting more analyses. 3 - An Analytical Framework for Intelligent Reviewer Recommendation Harry Wang, Association Professor, University of Delaware, 42 Amstel Ave, Newark, DE, 19716, United States of America, hjwang@udel.edu, Kunpeng Zhang, Sean Kilgallon Recruiting reviewers for academic conferences and journals is a daunting task for conference organizers and journal editors. In this paper, we propose an intelligent approach for reviewer identification based on techniques such as text mining, social network analysis, and recommender system. We collect data from online paper repositories and research social network sites. We evaluate our approach by conducting user studies and experiments. 4 - Semi-supervised Article Selection for Medical Systematic Reviews Jun Liu, Assistant Professor, Dakota State University, 422 SW 8th Street, Apt. 16, Madison, SD, 57042, United States of America, jun.liu@dsu.edu, Prem Timsina, Omar El-gayar We developed a semi-supervised learning based classifier to identify articles that can be included in of medical systematic reviews. Through an empirical study, we demonstrated that semi-supervised approach is a viable technique for selecting articles for systematic reviews when only a few number of training samples are available, and a combination of semi-supervised and active learning can further optimize the article selection process. Chair: Yan Huang, Assistant Professor, Stephen M. Ross School of Business, University of Michigan, 701 Tappan St. R5322, Ann Arbor, MI, 48109, United States of America, yphuang@umich.edu 1 - Who Hath the Crystal Ball? Antecedents of Advanced Analytics Adoption in Firms Ajit Sharma, Ross School of Business, 701 Tappan Street, Ann Arbor, MI, United States of America, asharmaz@umich.edu, Mayuram Krishnan, Yan Huang It is increasingly evident that the ability to gain forward-looking insight from advanced analytics will differentiate the winners from the losers. However, it is not clear what differentiates firms that are able to leverage these new technologies from those that do not. In this paper we attempt to answer this question by empirically investigating the antecedents of predictive analytics usage within firms. We present our findings and discuss managerial implications. 2 - Gain from Loss: Crowdsourcing Contests Bring Favorable Superstar Effect Shunyuan Zhang, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States of America, shunyuaz@andrew.cmu.edu, Param Vir Singh, Anindya Ghose We report favorable superstar effect, which contradicts with adverse superstar effect in Economic literature, and argue the unique crowdsourcing setting is the cause. Competing with superstars decreases one’s chance of winning but leads to an improved performance in the next competition. Contestants are self-selected to solve the trade-off problem. Our work suggests a promising crowdsourcing practice that generates spill-over effects and identifies participants who are more capable of learning. SA25 25-Room 402, Marriott Business Analytics and Innovation Sponsor: Information Systems Sponsored Session

SA24 24-Room 401, Marriott Data Mining for Decision Making Sponsor: Artificial Intelligence Sponsored Session

Chair: Iljoo Kim, Assistant Professor, Saint Josephís University, 347 Mandeville Hall, 5600 City Avenue, Philadelphia, PA, 19131, United States of America, ikim@sju.edu 1 - Rating Corporate Bonds using Deanfis Analysis – A Muli-modeling Approach Rashmi Malhotra, Associate Professor, Saint Joseph’s University, 5600 City Avenue, Philadelphia, PA, 19131, United States of America, rmalhotr@sju.edu, Davinder Malhotra This paper illustrates the use of data envelopment analysis (DEA) and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) to differentiate between bonds with credit ratings. This study measures the relative performance of corporate bonds using DEA. The ANFIS model creates a rule-based system that can aid the decision-maker in making decisions regarding the implications of a decision. This study proposes a modeling technique that jointly uses the two techniques to benefit from the two methodologies.

45