Community Detection in Large Social Networks Based on Relationship Density

Numerous algorithms have been proposed for detecting underlying community structure in social network. However, the majority of existing methods focus on modularity or structure similarity, and the effectiveness on complicated networks is still far from satisfactory. In this paper, we propose a novel community detection algorithm on the basis of newly defined relationship density in social networks. According to discrepant influence of each vertex, core and auxiliary vertices are introduced to represent the users who influence others significantly and who have a negligible impact, respectively. To summarize, the proposed method consists of two stages: 1) core vertices are clustered into kernel communities in the decreasing order of relationship density and 2) auxiliary vertices are assigned into the closest connected community formed in the previous stage. Experiments on three real networks are conducted to validate the proposed community detection algorithm based on relationship density (CDRD), which achieves appreciable performance improvement over other baseline methods in terms of F1-score.