Informs Annual Meeting 2017

MA31

INFORMS Houston – 2017

4 - Supplier Cost Reduction: Competition and Endogenous Information Asymmetry Cuihong Li, University of Connecticut, School of Business, 2100 Hillside Road, Storrs, CT, 06269, United States, cuihong.li@uconn.edu Consider a hold-up problem in which suppliers exert cost-reduction efforts before the buyer offers contracts. We study the impact of supplier competition and endogenous information asymmetry generated by mixed strategy of suppliers. 350E Machine Learning & Data Mining Sponsored: Artificial Intelligence Sponsored Session Chair: Ravi Aron, Johns Hopkins University, Baltimore, MD, 21202, United States, raviaron@jhu.edu 1 - Predicting Complex Outcomes using Machine Learning: A Dimensionality Reduction Approach Praveen Pathak, Warrington College of Business, University of Florida, Gainesville, FL, 32611, United States, praveen.pathak@warrington.ufl.edu Machine learning techniques are useful in many business domains and specifically to predict outcomes in big data. However big data inherently has a variety of features. The efficacy of machine learning algorithms might be reduced give these many features. But machine learning techniques can be used to do feature selection out of these myriad of features. Such feature selection leads to reduction in dimensionality of features. We use machine learning techniques for dimensionality reduction and show how complex outcomes could be predicted. 2 - Predicting Safe Handoffs, using Data Mining to Determine the Factors that Predict Safer Hand off Behaviors in Inpatient Settings Phillip Phan, Alonzo and Virginia Decker Professor, Johns Hopkins Carey Business School, Baltimore, MD, 21202, United States, pphan@jhu.edu, Soo-Hoon Lee This study seeks to discover a multi-level recursive model of organizational and process factors that determine the adoption and use of electronic hand-off tools in clinical settings. 3 - So Far So Good? Predicting Patient State Transition in Chronic Care using Machine Learning Ravi Aron, Johns Hopkins University, Carey Business School, 100 International Drive, Baltimore, MD, 21202, United States, raviaron@jhu.edu We address the problem of predicting state transition in chronic diseases. We use techniques drawn from Machine Learning - in specific ensemble learning techniques - to predict a patient’s likelihood of state transition. We contrast our approach with traditional estimation models and find that our method based on ensemble learning outperforms these methods. 350F Social Network Analytics Invited: Social Media Analytics Invited Session Chair: Peng Xie, Georgia Institute of Technology, Atlanta, GA, 30309, United States, peng.xie@scheller.gatech.edu 1 - A Deep Look into Helpful Online Physician Reviews Lina Zhou, University of Maryland-Baltimore County, 1000 Hilltop Circle, Baltimore, MD, 21250, United States, zhoul@umbc.edu, Nujood Alodadi Compared with online reviews of other types of products, physician reviews have been much less studied. Much remains to understand what constitutes a helpful online physician review. A few extant studies are focused on analyzing word-level features by leveraging general purposed lexicons. These methods are unable to address domain-specific characteristics of physician reviews such as embedded semantics. They also overlooked social network features of online reviews. This research is aimed to advance our knowledge about helpful online physician reviews by incorporating semantics and social network information. MA29 MA30

2 - Revealing Puzzles of the Effect of Online Brand Community Participation on Purchase Frequency Xuehua Liao, Ph.D, Sun Yat-sen Business School, Guangzhou, China, liaoxueh@mail2.sysu.edu.cn, Kang Xie, Jinghua Xiao Numerous studies on how online brand community engagement affects purchase generate mixed findings, but slim evidence deep into the contradiction. This study tries to reveal it from perspective of uncertainty reduction theory. Realistic engagement and purchase behavior data are collected to test our hypotheses. Results show that whether users reduce their uncertainty about the firm or product through communication with others plays an important role. This study enriches uncertainty reduction theory by extending its application condition from bilateral to trilateral relation. Also, it puts forward a fresh analytical perspective to shed light on the engagement-purchase effect confusion. 3 - Strategic Information Dissemination in Healthcare Workers Network Xiaowei Mei, University of Florida, 1454 Union Road, #11, Gainesville, FL, 32603, United States, xmei@ufl.edu How adoption of new healthcare management practices among healthcare workers can be promoted by disseminating information through their work related network? We examine this phenomenon by studying the adoption of a new medicine in a large hospital in the US. We construct a comprehensive work related network of doctors and other supporting healthcare staff based on their past collaboration over one year period. Then we examine the diffusion of new medicine in this network over time. We find that good health care management practices can be promoted by strategically seeding information about them in the work related network of health care workers. 4 - A Novel Literature Retrieval Model Based on Wisdom of Crowds in the Citation Network Ruiyun Xu, City University of Hong Kong, Kowloon Tong, Hong Kong, ruiyunxu2-c@my.cityu.edu.hk, Hailiang Chen, J. Leon Zhao We develop an innovative retrieval model by leveraging the wisdom of crowds embedded in the citation network to enable a fast and comprehensive article search. Instead of keywords, an abstract is used as query input to better capture a user’s information need. To assess the relevance between the query and each article in the corpus, we borrow from the research community’s aggregate opinion by looking at whether other articles that are topically similar to the query cite the article or not. We conduct experiments on the dataset collected from the three leading IS journals (ISR, JMIS, and MISQ) from 1977 to 2016. Our retrieval model has achieved better performance compared with the baseline models. 5 - Network Structure and Predictive Power of Social Media for the Bitcoin Market Following the recent discovery of social media’s predictive power for financial markets, we try to advance the literature by evaluating the role of social media network structure in distinguishing between value-relevant information and noises. Using data from the Bitcoin market, we provide empirical evidence that loosely-connected social media discussion networks are more accurate in predicting future returns. Although social media information linkages cause information free riding and damage the overall network prediction accuracy, they nevertheless serve as landmarks for identifying informed social media participants. 351A Academic Job Search Panel Invited: INFORMS Career Center Invited Session Chair: Warren Hearnes, Cardlytics, General Session Session, Atlanta, GA, 30308, United States, whearnes@hotmail.com 1 - Academic Job Search Warren Hearnes, Cardlytics, Atlanta, GA, whearnes@hotmail.com This panel discusses the academic interview process and do’s and don’ts associated with the job search. In addition to comments by current and former search chairs, time will be provided for questions and answers. Peng Xie, Georgia Institute of Technology, Room 907, 100 10th Street, Atlanta, GA, 30309, United States, peng.xie@scheller.gatech.edu MA31

143