Modeling Social Networks through User Background and Behavior

We propose a generative model for social networks, both undirected and directed, that takes into account two fundamental characteristics of the user: background (specifically, the real world groups to which the user belongs); and behavior (namely, the ways in which the user engages in surfing activity and occasionally adds links to other users encountered this way). Our experiments show that networks generated by our model compare very well with data from a host of actual social networks with respect to a battery of standard metrics such as degree distribution and assortativity, and verify well known predictions about social networks such as densification and shrinking diameter. We also propose a new metric for social networks intended to gauge the level of surfing activity, namely the correlation between degree and Page rank.