Influenza prediction from social media texts using machine learning

Abstract

The monitor and prediction of influenza through conventional modes seem obsolete in the Internet era. The flu prediction using machine learning methods is more resourceful for both economically as well as logistically, than traditional methods. The machine learning methods exploit freely available data on social media platforms. The main objective of this paper is to harness the power of publicly available data through machine learning methods. This work involves the scraping of textual data from six social media platforms and training of five machine learning predictors. The paper also proposes a three-layered prediction method based on support-vector-regression for predicting trends of influenza spread. The results obtained from proposed model are validated and analysed against other four machine learning methods. The results reveal that proposed model has better predictive accuracy in terms of least prediction errors among the five methods. Thus, it can provide effective means to control and to prevent the flu outbreak.