Social media prediction: a literature review

Social Media Prediction (SMP) is an emerging powerful tool attracting the attention of researchers and practitioners alike. Despite its many merits, SMP has also several weaknesses, as it is limited by data issues, like bias and noise, and the lack of confident predictions and generalizable results. The goal of this paper is to survey popular and trending fields of SMP from 2015 and onwards and discuss the predictive models used. We elaborate on results found in the literature, while categorizing the forecasting attempts, based on specific values (source of data, algorithm used, outcome of prediction etc.). Finally, we present our findings and conduct statistical analysis on our dataset and critique the outcome of the attempted prediction reported by the reviewed papers. Our research indicates that results are ambiguous, as not all forecasting models can predict with high accuracy, and prediction seems dependable on the associated field, although some of the documented attempts are promising. More than half (53.1%) of the examined attempts achieved a valid prediction, nearly one fifth (18.8%) did not, while the remaining 28.1% is characterized as plausible or partially validated. By reviewing recent and up-to-date literature and by providing statistics, this paper provides SMP researchers with a guide on methods, algorithms, techniques, prediction success and challenges on three main categories that aid SMP exploration.