Temporal interaction biased community detection in social networks

Community detection in social media is a fundamental problem in social data analytics in order to understand user relationships and improve social recommendations. Although the problem has been extensively investigated, most of the research examined communities based on static structure in social networks. Our findings within large social networks such as Twitter, show that only a few users have interactions or communications within any fixed time interval. It is not difficult to see that it makes more potential sense to find such active communities that are biased to temporal interactions of social users, rather than relying solely on static structure. Communities detected with this new perspective will provide time-variant social relationships or recommendations in social networks, which can greatly improve the applicability of social data analytics. In this paper, we address the proposed problem of temporal interaction biased community detection using a three-step process. Firstly, we develop an activity biased weight model which gives higher weight to active edges or inactive edges in close proximity to active edges. Secondly, we redesign the activity biased community model by extending the classical density based community detection metric. Thirdly, we develop two different expansion-driven algorithms to find the activity biased densest community efficiently. Finally, we verify the effectiveness of the extended community metric and the efficiency of the algorithms using three real datasets.