An Incremental Approach for Hierarchical Community Mining in Evolving Social Graphs

Community members which are highly connected with each other inside a community tends to create sub-communities, commonly termed as intrinsic or hierarchical communities. Finding intrinsic communities help us to reach out specific user needs, understanding the network dynamics and unveiling the functional and hidden aspects in the network, which is difficult without unveiling intra and inter-community all kinds of relationship. With the passage of time, members of a community may acquire different interests, leads to movement of members within different communities. Frequent changes in the relationship of members towards a community make the task of community detection even more challenging. In this work, we propose a new community detection method, embedded communities from evolving networks (ECEnet), for handling intrinsic communities in evolving networks. We adopt a density variation concept to detect the intrinsic communities in growing networks. We use a new membership function to measure the contiguity of a member towards a community. We use both synthetic and real-world social networks for our experimentation. Experimental results reveal that ECEnet is successful in detecting intrinsic or hierarchical communities in a dynamic scenario.