2015 Informs Annual Meeting

TC06

INFORMS Philadelphia – 2015

TC04 04-Room 304, Marriott Social Media Analytics Best Papers Finalist Competition Cluster: Social Media Analytics Invited Session

In this paper, we undertake a content analysis of the news feed regarding big data analytics. We develop a corpus of related news items. We code the corpus and prepare it for analysis by the Natural Language Processing (NLP) algorithms. Our purpose is to conduct a systematic analysis of the news contents to determine the primary themes being projected by the proponents of big data. We will explore ways to identify real value content that may help to draw meaningful inferences. 2 - Modeling Message Dissemination in Social Media for Emergency Communication Justin Yates, Professor, Francis Marion University, 4822 E Palmetto St, Florence, SC, United States of America, jyates@fmarion.edu, Xin Ma We introduce the Single-network Multi-message Social Media Message Dissemination Problem (SM-SMMDP) as a discrete optimization model to examine message dissemination in social media and to explore message dissemination strategies for government and non-government emergency management agencies. We implement a detailed computational experiment on representative small-scale networks and discuss implications for real-world application. 3 - Analyzing and Predicting Threatening Language in #gamergate Tweets Cheyanne Baird, Senior Linguistic Specialist, SAS Institute Inc., Boston Regional Office, Prudential Tower, 800 Boylston St., Suite 2200, Boston, MA, 02199, United States of America, Cheyanne.Baird@sas.com, Michael Wallis, Praveen Lakkaraju Aggressive language in online communities can thrive, escalate, and signal palpable threat. This is apparent in #gamergate tweets. Using SAS Text Analytics, we will show how language transitions from negative sentiment to threatening language in #gamergate tweets, building a model to predict probable threat in social media. TC06 06-Room 306, Marriott Dynamics and Information in Commodity Markets Sponsor: Financial Services Sponsored Session Chair: Richard Sowers, Professor, University of Illinois at Urbana- Champaign, Urbana, IL, 61801, United States of America, r-sowers@illinois.edu 1 - Model Uncertainty in Commodity Markets Sebastian Jaimungal, University of Toronto, Department of Statistical Sciences, Toronto, Canada, sebastian.jaimungal@utoronto.ca, ¡lvaro Cartea, Zhen Qin This paper studies the effect that model ambiguity in commodities have on the value of derivative contracts. The base model consists of three drivers: a mean- reverting diffusion, a mean-reverting jump, and a stochastic volatility component. We allow the agent to consider a wide class of alternate models, and penalize the differing components of the model individually. We demonstrate how agents alter their behavior in the presence of ambiguity and how derivatives and spreads are affected by it. 2 - Index Tracking with Futures Brian Ward, Columbia University, New York, NY, bmw2150@columbia.edu, Tim Leung Exchange Traded Funds (ETF) market continues to grow. Many ETFs are designed to track an index, and their leveraged benchmarks (2x, 3x, etc.) Since many of these indices are not directly traded, we consider tracking them using futures of various maturities. We do so both statically and dynamically, and backtest our strategies with empirical data. 3 - Investment in Commodities ETFs and Management of Contango Andrew Papanicolaou, NYU Polytechnic, 6 Metrotech Center, Brooklyn, Ny, 11201, United States of America, apapanic.brown@gmail.com The last two decades have seen growing investment in commodities markets. Commodities ETFs are popular but not a passive investment, as they will post losses in contango markets. The focus of this talk will be storable commodities, where uncertainty in the convenience yield reduces the Sharpe ratios. Losses are seen as an information premium, which is quantified through a Merton-type control problem.

Chair: Theodore Allen, Associate Professor, The Ohio State University, 1971 Neil Avenue, 210 Baker Systems, Columbus, OH, 43221, United States of America, allen.515@osu.edu 1 - Social Media Analytics: The Effectiveness of Marketing Strategies in Online Social Media Vilma Todri, PhD Candidate In Information Systems, NYU, 44 W 4th St, KMC Room 8-181, New York, NY, 10012, United States of America, vtodri@stern.nyu.edu, Panagiotis Adamopoulos This paper studies a novel social media venture and seeks to understand the effectiveness of marketing strategies in social media platforms by evaluating their impact on participating brands. We use a real-world data set and employ a promising research approach combining econometric with predictive modeling techniques. 2 - Predicting Crowd Behavior with Big Public Data Nathan Kallus, MIT, 77 Massachusetts Ave., E40-149, Cambridge, MA, 02139, United States of America, kallus@mit.edu We present efforts to predict the occurrence, specific timeframe, and location of crowd actions before they occur based on public data collected from over 300,000 open content web sources in 7 languages, from all over the world, ranging from mainstream news to government publications to blogs and social media. 3 - Participation vs. Effectiveness of Paid Endorsers in Social Advertising Campaigns: A Field Experiment Jing Peng, The Wharton School, University of Pennsylvania, 3730 Walnut Street Suite 500, Philadelphia, PA, 19104, United States of America, jingpeng@wharton.upenn.edu, Christophe Van den Bulte We investigate the participation and effectiveness of paid endorsers in viral-for- hire social advertising. Specifically, we investigate (i) how financial incentives affect potential endorsers’ participation and effectiveness in generating online engagements (likes, comments, and retweets), and (ii) which network characteristics and prior engagement characteristics are associated with participation and effectiveness. We conduct a large scale field experiment with an invitation design in which we manipulate both financial incentives and the soft eligibility requirement to participate in the campaign. The latter provides a strong and valid instrument to separate participation from outcomes effects. Since likes, comments, and retweets are count variables, and since potential endorsers can self-select to participate in multiple campaigns we ran, we use a Poisson lognormal model with sample selection and correlated random effects to analyze variations in participation and effectiveness. There are three main findings. (1) Financial incentive did not affect either participation or effectiveness. (2) Potential endorsers more likely to participate are often less effective. (3) Which characteristics are associated with effectiveness depends on whether success is measured in likes, comments, or retweets. Our findings provide new insights on how marketers can improve social advertising campaigns by better targeting and incenting potential endorsers. 4 - A Visual Monitoring Technique Based on Importance Score and Twitter Feeds Zhenhuan Sui, Graduate Research Assistant, The Ohio State University, 2501 Gardenia Drive, Columbus, Oh, 43235, United States of America, sui.19@osu.edu, David Milam, Theodore Allen We propose a visual monitoring technique based on topic models, a point system, and Twitter feeds for monitoring. The method generates a chart showing the important and trending topics that are discussed over a given time period which is illustrate the methodology using cyber-security cases. TC05 05-Room 305, Marriott Social Media Impact Cluster: Social Media Analytics Invited Session Chair: Les Servi, The MITRE Corporation, 202 Burlington Road, Bedford, MA, United States of America, lservi@mitre.org 1 - Big Data Means Big PR: A Review of News Coverage of Big Data in the Popular Press

Amir Gandomi, Assistant Professor, Ryerson University, 350 Victoria Street, Toronto, ON, M5B 2K3, Canada, agandomi@ryerson.ca, Murtaza Haider

319