Excavating social circles via user interests

The rapid evolution of modern social networks motivates the design of networks based on users’ interests. Using popular social media such as Facebook and Twitter, we show that this new perspective can bring more meaningful information about the networks. In this paper, we model user-interest-based networks by deducing intent from social media activities such as comments and tweets of millions of users in Facebook and Twitter, respectively. These interactive contents derive networks that are dynamic in nature as the user interests can evolve due to temporal and spatial activities occurring around the user. To excavate social circles, we develop an approach that iteratively removes the influence of the communities identified in the previous steps by widely used Clauset, Newman, and Moore (CNM) community detection algorithm. Experimental results show that our approach can detect communities at a much finer scale compared to the CNM algorithm. Our user-interest-based model and community extraction methodology together can be used to identify target communities in the context of business requirements.