Title: Predictive Analytics with Big Social Data: Taking Stock and Looking Ahead
Authors: Niels Buus Lassen, Lisbeth la Cour, and Ravi Vatrapu
Abstract: Current research projects at the Computational Social Science Laboratory (http://cssl.cbs.dk), Copenhagen Business School are addressing an academic research gap and real-world industry need to describe, model, analyse and explain large-scale interactions on organisations’ social media channels as individuals' associations to ideas, values, identities etc. Towards this end, we are developing and evaluating a set-theoretical approach to big data analytics termed “Social Set Analysis” (SSA). SSA consists of three primary research activities: (a) theorising, modelling, and collecting big social data about organisations (e.g., Danish Cancer Society’s official Facebook page); (b) combining those big social data sets with in-house organisational data sets (e.g., Customer Relationship Management systems); and finally (c) analysing the combined datasets by applying set theoretical methods and tools (crisp sets, fuzzy sets, rough sets, random sets and Bayesian sets).
This talk will outline the SSA approach and provide an overview of the extant literature on predictive analytics with social media data. First, we provide an outline of the set theoretical approach to computational social science. Second, we discuss the difference between predictive vs. explanatory models and the scientific purposes for and advantages of predictive models. Third, we present and discuss the foundational statistical issues in predictive modelling in general with an emphasis on social media data. Fourth, we present a selection of papers on predictive analytics with social media data and categorize them based on the application domain, social media platform (Facebook, Twitter etc.), independent and dependent variables involved, and the statistical methods and techniques employed. Fifth and last, we offer some reflections on predictive analytics with social media data.