Quantifying the evolutions of social interactions

Abstract

Understanding social interactions and their evolutions has important implications for exploring the collective intelligence embedded in social networks. However, the dynamic patterns in social interactions are not well investigated, and the time when the interactions take place is ignored in the existing studies. In this paper, a graph-based model incorporating a decay function is proposed to study the evolutions of social interactions quantitatively. In the experiments, the proposed model is applied to Digg dataset. The results show that the node degree of the social interaction graph follows the power law distribution, and the users' interactions have locality property. Furthermore, the results demonstrate that the evolutions of social interactions are useful for tracking the trends of topics. © 2014 Springer International Publishing Switzerland.