Detecting communities in online social networks

A community can be defined as a subset of the users in a social network that is more tightly interconnected than the overall network. Communities are useful, for instance, to guide information dissemination and acquisition, to recommend or introduce people who would likely benefit from direct interaction, and to express access control policies. In this paper, we study algorithms for automatically detecting communities in a social network. Using data from a university social network, we show that individual users are typically part of several communities, such as communities based on dormitory, matriculation year, or department. We show that existing algorithms for detecting communities associate each user with only one of the relevant communities and, consequently, fail to detect the multiple communities that exist in the network. Finally, we present and evaluate a new algorithm that can detect a specific community in the social network with high accuracy when given a small subset of the users in that community.