Top-L Most Influential Community Detection over Social Networks (technical Report)

In many real-world applications such as social network analysis and onlinemarketing/advertising, the community detection is a fundamental task toidentify communities (subgraphs) in social networks with high structuralcohesiveness. While previous works focus on detecting communities alone, theydo not consider the collective influences of users in these communities onother user nodes in social networks. Inspired by this, in this paper, weinvestigate the influence propagation from some seed communities and theirinfluential effects that result in the influenced communities. We propose anovel problem, named Top-L most Influential Community DEtection (TopL-ICDE)over social networks, which aims to retrieve top-L seed communities with thehighest influences, having high structural cohesiveness, and containinguser-specified query keywords. In order to efficiently tackle the TopL-ICDEproblem, we design effective pruning strategies to filter out false alarms ofseed communities and propose an effective index mechanism to facilitateefficient Top-L community retrieval. We develop an efficient TopL-ICDEanswering algorithm by traversing the index and applying our proposed pruningstrategies. We also formulate and tackle a variant of TopL-ICDE, nameddiversified top-L most influential community detection (DTopL-ICDE), whichreturns a set of L diversified communities with the highest diversity score(i.e., collaborative influences by L communities). We prove that DTopL-ICDE isNP-hard, and propose an efficient greedy algorithm with our designed diversityscore pruning. Through extensive experiments, we verify the efficiency andeffectiveness of our proposed TopL-ICDE and DTopL-ICDE approaches overreal/synthetic social networks under various parameter settings.