Recurrent random forest for the assessment of popularity in social media

Popularity in social media is mostly interpreted by drawing a relationship between a social media account and its followers. Although understanding popularity from social media has been explored for about a decade, to our knowledge, the extent to which the account owners put efforts to enhance their popularity has not been evaluated in detail. In this paper, we focus on Twitter, a popular social media, and consider the case study of the 2016 US elections. More specifically, we aim to assess whether candidates endeavor to improve their style of tweeting over time to be more attractive to their followers. An ad hoc-defined predictive model based on a recurrent random forest is used for this purpose. To this end, we build a classification model whose features are obtained from the characteristics of a set of content/sentiment information extracted from the tweets. Next, we derive an index of social media popularity for both candidates. Results show that Trump wisely exploited Twitter to attract more people by tweeting in a well-organized and desirable manner and that his tweeting style has increased his popularity in social media. The differences in the tweeting styles of the two presidential candidates and the links between the sentiments arising from candidates’ tweets and their popularity index are also investigated.