Some Advances of Statistical Methodologies for Community Detection in Social Networks

Abstract

The growth in the big data regime and the popularity of new media have enhanced a research migration to network pattern recognition and analysis. In the real world, society offers a wide variety of possible communities, such as schools, families, and firms. For this reason, community detection draws much attention as it is important to many applications in business and social sciences. Since communities are mainly related to the edges and distances among vertices, we first introduce how to efficiently glimpse a large network by a graph partition tool. To provide statistical significance for both structure and attribute communities in social networks, we generalize the scan statistics to accommodate the network features. Simulation studies show that the scan statistics provide better detection performances, and empirical studies show the differences among the proposed methods and the previous methods. Keywords: Social Networks; Community detection; Graph Partition; Scan Statistic.