Statistical analysis and anomaly detection of SMS social networks

Social network analysis has attracted intensive interests by researchers from multiple disciplines. However most of the existing work is descriptive nature, and statistical network analysis remains an active area of research. In this paper, we model and study two facets of the social networks in short message services (SMS). One is the structure of the contact networks of mobile users, the other is users' messaging behavior pattern. We want to account for the heterogeneity in behavior so that to identify abusive usage such as spamming through the study. We use power-law mixture model to capture community formation behaviors, the first facet, and use Poisson-panel mixture models to uncover abnormal behaviors in text messaging. Our results show heterogeneity of the consumers' sending behavior, also there are two major types of community formation behavior in SMS network.