Informs Annual Meeting 2017

SD31

INFORMS Houston – 2017

SD29

2 - Using Microblogging to Predict IPO Success Dirk Van den Poel, Professor of Data Analytics/Big Data, Ghent University, Faculty of Economics and Business Administration,

350E Healthcare Informatics Sponsored: Artificial Intelligence Sponsored Session Chair: Haijing Hao, University of Massachusetts-Boston, 12 Inman Street Apt 31, Cambridge, MA, 02139, United States, haohaijing@gmail.com 1 - Social Media and Health Insurance Choice in China Haijing Hao, University of Massachusetts-Boston, 12 Inman Street Apt 31, Cambridge, MA, 02139, United States, haohaijing@gmail.com, Yunfeng Shi, Anqichen Shi China has been undertaking health insurance reforms in the past couple of decades, transforming from narrow coverage of government paid insurance to limited government staff to wider population from all walks of people. Chinese people have more and flexible choices of health insurance. This study examines the daily social media usage and health insurance choices in China based on a national survey study. 2 - Dynamics and Misinformation in Health Social Networks Wencui Han, University of Illinois at Urbana Champaign, 1206 S. Sixth Street, 350 Wohlers Hall, Champaign, IL, 61820, United States, wenhan@illinois.edu Health social networks provide patients opportunities to connect with each other. However, misinformation on these platforms poses threats to patient safety. This paper uses physician-evaluated posts collected from a popular health social network site to explore factors can be used to predict and monitor misinformation. We focus on the impacts of two main concepts that embedded in the dynamics of the interaction on social networks: similarity between responder and question poster, and dynamic clique variables. 3 - Digital Divide in Online Health Community Haijing Hao, University of Massachusetts-Boston, 12 Inman Street Apt 31, Cambridge, MA, 02139, United States, haohaijing@gmail.com The present study intends to explore what factors may have an impact on online health utilization in China by examining various economic-related variables. We use Tableau mapping features to illustrate the geographical visualization and also run quantitative fixed effect regression models. Our preliminary results show that percentage population that can access the Internet of a province in China has a statistical significant impact on the online health utilization of that province. 4 - Efficiently Identifying Trends in Healthcare using Text Mining, and Topic Detection and Tracking Yu-Hung Chiang, National Cheng Kung University, Tainan City, Taiwan, momois87@hotmail.com Most published healthcare research literature is recorded online databases such as PubMed. As the number of online articles increases rapidly, it becomes challenging for researchers to accurately and efficiently identify current research trends in specific topics of healthcare. In this study, we construct a trend analysis method to obtain research topics from PubMed. Using text mining methods and topic Detection and Tracking (TDT), we aim to improve the efficiency in identifying and tracking topics from large databases. 350F Social Media Use and New Product Development Performance Invited: Social Media Analytics Invited Session Chair: Debasish N Mallick, University of St. Thomas, Minneapolis, MN, 55403, United States, dnmallick@stthomas.edu Chair: Julie Zhang, University of Massachusetts Lowell, Lowell, MA, 01854, United States, juheng_zhang@uml.edu 1 - Social Media Use & New Product Development Performance Debasish N. Mallick, University of St. Thomas, Opus College of Business, 1000 La Salle Avenue # TM.H.443, Minneapolis, MN, 55403, United States, dnmallick@stthomas.edu Use of social media is becoming increasingly popular in NPD. Yet, the performance impact of social media use in NPD remains inconclusive. Using a cross industry survey of new product development projects, we explore the factors affecting the relationship between social media use and NPD performance SD30

Tweekerkenstraat 2, Gent, B.9000, Belgium, Dirk.VandenPoel@UGent.be, Michel Ballings

This research investigates the influence of microblogging - in particular Twitter social media messages - on Initial Public Offering (IPO) success. We analyze the (potential) impact of the number of tweets and their sentiment on (1) the difference between the IPO price and the closing price of the stock at the day of becoming public, and (2) the difference between the closing price on the first day of trading and the closing price after three months of trading.

SD31

351A George B. Dantzig Dissertation Invited: The George B. Dantzig Dissertation Prize Invited Session Chair: Wedad Elmaghraby, University of Maryland, R H Smith, College Park, MD, 2, United States, welmaghr@rhsmith.umd.edu 1 - Robust Optimization for Renewable Energy Integration in Power System Operations Álvaro Lorca, Pontificia Universidad Católica de Chile, Departamento de Ingeniería Eléctrica, Av Vicuna Mackenna 4860, Santiago, 7820436, Chile, alvarolorca@uc.cl Motivated by the increasing adoption of variable renewable energy sources such as wind and solar power, this talk will present robust optimization models and algorithms that determine innovative methods to operate modern power systems. In particular, we will present a multistage adaptive robust optimization model for the unit commitment problem, i.e. for the problem of scheduling power generation over one whole day, including on/off and dispatch decisions. The most innovative aspect of this model is that the causality of the dispatch process is respected, that is, dispatch decisions at any given time can only depend on the revealed uncertainty up to that time. We will also discuss solution methods based on cutting planes, the concept of dynamic uncertainty sets to account for spatial and temporal correlations in renewable power uncertainty, and the potential impact in the electric power industry. 2 - Deep Exploration via Randomized Value Functions Ian Osband, Stanford University, Stanford, CA, United States, ian.osband@gmail.com In this dissertation we present an alternative approach to deep exploration through the use of randomized value functions. Our work is inspired by the Thompson sampling heuristic for multi-armed bandits which suggests, at a high level, to ``randomly select a policy according to the probability that it is optimal’’. We provide insight into why this algorithm can be simultaneously more statistically efficient and more computationally efficient than existing approaches. We leverage these insights to establish several state of the art theoretical results and performance guarantees. Importantly, and unlike previous approaches to deep exploration, this approach also scales gracefully to complex domains with generalization. We complement our analysis with extensive empirical experiments; these include several didactic examples as well as a recommendation system, Tetris, and Atari 2600 games. 3 - Robust Solutions for Geographic Resource Allocation Problems Mehdi Behroozi, Northeastern University, Allston, MA, 02134, United States, m.behroozi@neu.edu The concept of geography is a fundamental one in many domains such as transportation, facility location, logistics, surveillance, reconnaissance, and territory division, among many others. This dissertation is devoted to the study of geographic resource allocation (GRA) problems. In such a problem, our goal is to minimize or maximize an objective function that is defined over a geographic region. The distinguishing attribute of such problems is that spatial properties such as distance, shape, perimeter, area, connectedness, fatness, or convexity will appear in these problems either in the objective function or as a constraint. This dissertation considers these problems under uncertainty and aims to achieve some sort of robustness against that uncertainty. In order to take advantage of the above spatial properties and geometric principles to overcome the drawbacks of current approaches to solving such problems, we adopt an interdisciplinary approach that combines concepts, theories, and methods from computational geometry, geometric probability theory, calculus of variations, infinite- dimensional optimization, and topology. We first show the inability of the existing operations research tools to obtain reasonable solutions for some special but common and practical cases, like clustered data, and then we show that our approach makes these problems easy to solve. The results of this research have been used by Oracle Labs in a routing package for their forthcoming spatial database products and by a major mapping company in a worldwide LIDAR surveying project to scan neighborhoods and collect street level data in five continents

111