A Centrality-Based Local-First Approach for Analyzing Overlapping Communities in Dynamic Networks

With the increasing demand of dynamic graph data analysis, mining communities in time-evolving data has been a research hotspot. However, traditional community detection methods have efficiency issue in the huge dynamic network data and rarely consider about overlapping communities. In this paper, we first propose a centrality-based local-first approach for overlapping community discovery in static network, called CBLF. Different with the traditional top-down approach, CBLF detects communities from central nodes and theirs neighbors which conforms to reality better. Then we present a novel evolutionary community detection approach called CBLFD based on this effective approach and sequence smoothing mechanism. Experimental results on real-world and synthetic datasets demonstrate that these algorithms achieve higher accuracy and efficiency compared with the state-of-art algorithms.