Revealing the relationship of topics popularity and bursty human activity patterns in social temporal networks

Abstract

In social networks, how human activity patterns affect the popularity of topics has always been the focus of research. In this paper, a quantitative temporal analysis of the dynamics of topics popularity in Sina Weibo system was provided. Firstly, the popularity time series of 1167 topics were clustered into four clusters by K-Spectral Centroid (K-SC) clustering algorithm. Secondly, for each cluster, we calculated the exponents of topic popularity decay distribution α and the exponents of inter-activity time distribution β, respectively. Two interesting results were found: one is that the peak fraction F of topics popularity positively correlated with the topics popularity decay exponent α; the other is that bursty activity patterns in social network significantly affect topics popularity dynamics: there is a positive correlation between exponent α and exponent β. Finally, we proposed an extended SI (susceptible–infected) epidemic model with incorporate bursty human activity and verified the results by simulation.