Informs Annual Meeting 2017

SB31

INFORMS Houston – 2017

SB31

3 - Fuzzification of Sorting for Multiple Combinatorial Optimization Applications

351A Nicholson Student Paper Competition II Invited: Nicholson Student Paper Prize Invited Session Chair: Cole Smith, Clemson University, 110 Freeman Hall, Clemson, SC, 29634, United States, jcsmith@clemson.edu Co-Chair: Hayriye Ayhan, Georgia Institute of Technology, School of Industrial & Systems Eng, 755 Ferst Dr NW, Atlanta, GA, 30332-0205, United States, hayhan@isye.gatech.edu Co-Chair: Sergiy Butenko, Texas A&M University, 4037 Emerging Technologies Building, Mail Stop 3131, College Station, TX, 77843- 3131, United States, butenko@tamu.edu 1 - Minimizing Multimodular Functions and Allocating Capacity in Bike-sharing Systems Daniel Freund, Cornell University, 109 Lake St, Ithaca, NY, 14850, United States, df365@cornell.edu We consider the strategic question of how to (re-)allocate dock-capacity in bike- sharing systems. We develop mathematical formulations for variations of this problem and exhibit discrete convex properties in associated optimization problems. This allows us to design a polynomial-time allocation algorithm that can handle constraints such as a limit on the number of docks moved in the system. We apply our algorithm to data sets from Boston, New York City, and Chicago. 2 - Learning Preferences with Side-information Andrew A. Li, Massachusetts Institute of Technology-ORC, 77 Massachusetts Avenue, Bldg. E40-149, Cambridge, MA, 02139, United States, aali@mit.edu Product and content personalization is now ubiquitous in e-commerce. Companies seek to use a variety of information on the interactions between a product and a customer to drive personalization decisions. We formalize this problem as one of recovering a large-scale matrix, with side information in the form of additional matrices of conforming dimension. We provide an efficient algorithm that is practical for massive datasets and demonstrate its performance on data from a music streaming service. 3 - Tight Weight-dependent Competitive Ratios for Online Edge- weighted Matching, with Application to Revenue Management Will Ma, Massachusetts Institute of Technology, Cambridge, MA, Contact: willma353@gmail.com Online bipartite matching is a classical problem in the study of online algorithms. These problems can be abstracted as follows: there are fixed resources that must be allocated on-the-fly, without knowledge of future demand. This talk studies the generalizations of these problems when each resource could be sold at multiple, known prices. We show how to generalize the algorithms to accommodate multiple prices, and achieve a tight competitive ratio under the integral or asymptotic settings. 4 - Computationof the Bootstrap: Complexity, Exact Algorithms, and Deterministic Approximations Bradley E. Sturt, Massachusetts Institute of Technology, Somerville, MA, Contact: bsturt@mit.edu The bootstrap is a randomized method for calculating quantities, such as confidence intervals, directly from data. Because variability from randomization can lead to inaccurate outputs, we propose a deterministic approach. We present computational complexity results for the bootstrap method and the first FPTAS for the bootstrap method. We demonstrate that the algorithms run fast in practice and produce bootstrap confidence intervals that are more accurate compared to traditional methods.

Paul Eugene Coffman, Technical Leader, Ford Motor Company, 6100 Mercury Drive, Dearborn, MI, 48126-2746, United States, gcoffman@ford.com, Stephany Coffman-Wolph Many efficient heuristics for combinatorial algorithms (e.g., bin packing, knapsack, and queueing) rely on fast sorting algorithms. Using an established three-step framework for the fuzzification of algorithms, sorting algorithms can be converted into fuzzy sort algorithms. These “fuzzy” sort algorithms often provide computational improvements with minimal acceptable trade-offs, yielding “good enough” solutions. Extending selected traditional heuristic algorithms with fuzzy sorting should generate acceptable solutions faster than the base algorithms for these problems types. 350F Social Media Analytics for Healthcare Improvement I Invited: Social Media Analytics Invited Session Chair: Shu He, University of Illinois at Urbana-Champaign, Urbana, IL, 61802, United States, shuhe2@illinois.edu 1 - Big Data Analytics for Text-based Data in Healthcare: A Review and Research Directions Priya Nambisan, University of Wisconsin - Milwaukee, Northwest Quadrant Building B, Rm #6410, 2025 East Newport Avenue, Milwaukee, WI, 53201, United States, nambisap@uwm.edu Many recent studies have focused on developing algorithms for analyzing text based health data such as patient postings on social media, physician notes, etc. However, it is not clear whether these studies are contributing to healthcare research or to research on algorithm development. In this study, 40 papers in this topic were systematically analyzed to evaluate the effectiveness of the algorithms to analyze text based data and to determine the relevance of the findings for the healthcare domain. Based on the analysis, several limitations of these studies are identified and future research directions are discussed. 2 - The Effectiveness of Social Media-enabled Patient Communities on Health Goal Attainment: An Approach of Survival Analysis Jiahe Song, Assistant Professor, Western Michigan University, Kalamazoo, MI, United States, carrie.song@wmich.edu, Pei Xu This study explores how social media-enabled patient communities affect health goal attainment. Applying social cognitive theory, we studied the online antecedents of goal attainment from the respects of social support and self- reflection. A survival analysis was applied to a unique dataset of patients’ interactions and goal progress. Findings indicate that the two social support factors, informational and emotional support, have opposite effects (direct and buffering) on goal attainment. The two self-reflection factors, verbal and leisure update, also affect the possibility of goal attainment oppositely. The research provides important implications to online disease self-management. 3 - Expanding the Reach of Humanitarian Organizations on Social Media Eunae Yoo, Arizona State University, Tempe, AZ, United States, eunae.yoo@asu.edu, Elliot Rabinovich, Bin Gu Humanitarian organizations have been using social media to broadcast important messages to their stakeholders. On social media, a user’s follower count typically represents the user’s audience size, or reach. We explore the mechanisms that drive follower growth for organizations involved with disaster relief using a large data set from Twitter during the 2016 Ecuador earthquake. Our data capture the dynamic nature of the user network one week before and after the disaster, and we find that the follower bases of the organizations in our sample increase dramatically after the earthquake. 4 - The Nature of Connections on Twitter and its Dynamics Youngsoo Kim, Singapore Management University, School of Information Systems, 80 Stamford Road, Singapore, 178902, Singapore, yskim@smu.edu.sg Twitter is one of the leading social network sites (SNS). Prior studies have mainly focused on social networking behavior or information diffusion process on Twitter. By contrast, this study explores (1) the network formation (link connection with the focus on the network structural format), (2) the nature of link (online versus offline relationship) and its impact on communication frequency, and (3) the dynamics of homophily and interaction over time. SB30

45