Disentagling Social Networks Inferred from Call Logs

M. Cebrian, A. Pentland, S. Kirkpatrick


Analysis of an unusually detailed telephone call data set — a month of nearly all mobile and landline phone calls placed during August 2005 the United Kingdom — allows us to identify several different types of social networks that are formed, and relate them to different activities that generate them. We distinguish, among others, work-related and personal or leisure-focused activities and show that the networks they form have very different characteristics. Our principal tool for the analysis, k-core decomposition, shows that distinct distributions of connectivity are present in the two spheres, and that this differentiation affects dramatically the dynamics of information diffusion. Both differ from the simpler and more globally connected structure evident in communications data such as the Internet AS graph.

Related Content