Detecting Communities Using Social Ties

Many internet-based applications such as social networking websites, online viral marketing, and recommendation network based applications, use social network analysis to improve performance in terms of user-specific information dissemination. The notion of community in a social network is a key concept in such analyses and there has been significant work recently in identifying communities within a social network. In this paper, we formally define the notion of strength of a link, which was informally introduced by Granovetter, and present a divisive hierarchical clustering method to divide the nodes of a social network into disjoint communities. We also introduce the notion of clustering coefficient as a measure of the quality of a community or cluster. Our experimental results using some well-known benchmark social networks show that our method determines communities with better clustering coefficient than the well known Girvan-Newman method.