Hawkes Point Processes for Social Media Analytics

Abstract

Online social networks (OSNs) produce a huge volume of content and clickstream data over time as a result of continuous social interactions between users. Because these social interactions are not fully observable, the mining of such social streams is more complex than traditional data streams. Stochastic point processes, as a promising approach, have recently received significant research attention in social network analysis, in attempts to discover latent network structure of online social networks and particularly understand human interactions and behavior within the social networks. The objective of this paper is to provide a tutorial to the point process framework and its implementation in social media analytics. It begins by providing a quick overview of the history of Hawkes point processes as the most widely used classes of point process models. We identify various capabilities and attributes of the Hawkes point processes and build a bridge between the theory and practice of point processes in social network analytics. Then the paper includes a brief description of some current research projects that demonstrate the potential of the proposed framework. We also conclude with a discussion of some research opportunities in online social network and clickstream point process data.