Multi-Label Propagation for Overlapping Community Detection Based on Connecting Degree

Introduction With the growth of social media, social network analysis draws a great attention and becomes a hot research topic in the field of complex network, web mining, information retrieval, etc. An important aspect of social networks analysis is community structure (Newman, 2003). In general, community detection methods are classified into two categories: overlapping methods (and non-overlapping methods (Hofman & Wiggins, 2008)). The former allows communities overlap, while the latter assumes that a network only contains disjoint communities. In this paper, we focus on the overlapping community detection. To find overlapping community, researchers use a wide variety of techniques, such as Clique Percolation Method, COPRA (Gregory, 2010), etc. COPRA is very fast, but the result of COPRA is nondeterministic, so we propose an improved COPRA with high determinacy in this paper.