Hawkes processes for understanding heterogeneity in information propagation on Twitter

Abstract

Social platforms make information propagation anywhere anytime. Large quantity data recording information spreading is available for further understanding the intrinsic mechanism within these stochastic processes. Based on the empirical spreading trees of tweets on Twitter, the heterogeneity of Twitter users is explored, turning out the burstiness in the spreading process. A few super spreaders can significantly change the trends of information spreading. Moreover, an improved Hawkes process is designed in this study to better investigate users’ heterogeneity during information propagation. Verification is provided for accuracy and stability of the improved Hawkes model in simulating propagation patterns revealed in empirical sequential data, predicting spreading trends, and predicting probability of information outbreaks. Our improved Hawkes model is an effective spreading model for detecting and quantifying super spreaders during the propagation process, which may shed light on the control and prediction of information spreading in social media.