Fast detection of overlapping communities in online social networks

In today's world, social media networks capture interactions among people through comments on blogs, posts and feeds. The public availability of these networks has allowed researchers and businesses alike to delve more into these preferences so as to extract communities of similar interests which define their formation. Although community detection has been well applied to social networks, only recently has the need been felt for detecting overlapping communities. As people tend to have more than one preference over different products, it makes it difficult to put them in a single community. Also, the volume and scale of social media data makes it difficult to detect communities in real time which calls for faster implementations of existing algorithms. In this paper we first describe an existing algorithm which applies a game theoretic approach to determine overlapping communities within networks and show how a parallel implementation of the algorithm can be used to detect communities in lesser time than its previous implementations. We validate our implementation, by running experiments on some real world on-line social networks. We conclude by suggesting efficient ways to implement faster algorithms and topics of further research to detect and analyze social networks.