Generalization of Information Spreading Forensics via Sequential Dependent Snapshots

Abstract

Learning the characteristics of information spreading in networks is crucial in communication studies, social network sentiment analysis and epidemic investigation. Previous work on information spreading has been focused on the information source detection using either a single observation, or multiple but "independent" observations of the underlying network while assuming information spreads at a "uniform spreading rate" . In this paper, we conduct the first theoretical and experimental study on information spreading forensics, and propose a framework for estimating information spreading rates, information source start time and location of information source by utilizing "multiple sequential and dependent snapshots" where information can spread at different rates. We prove that our estimation framework generalizes the rumor centrality [1], and we allow heterogeneous information spreading rates on different branches in d -regular tree networks. We further show that our framework can provide highly accurate estimates for the information spreading rates on different branches, the source start time, and more accurate estimate for the location of information source than rumor centrality and Jordan center in both synthetic networks and real-world networks (i.e., Twitter).