Revealing the role of node similarity and community merging in community detection

Abstract

Many real-world networks are characterized by well-defined statistics of overlapping and nested communities. This paper proposes a series of seed-extension-based, overlapping, community detection algorithms to reveal the role of node similarity and community merging in community detection. First, we introduce core similarity to improve the definitions of degree centrality and fitness. Second, we propose a series of algorithms based on a strong and a weak community merging method. Finally, we design experiments to explore the effect of node similarity and community merging on community detection. Experimental results on artificial and real-world networks show that node similarity helps to identify communities when the community structure is easily identified but exacerbates the misclassification of nodes when the community structure is difficult to identify. Furthermore, community merging helps to maintain the integrity of the community when the community structure is easily identified, but merging leads to a community over-merger problem when the community structure is difficult to identify, especially for weak merging methods.