Detecting Community Structure in Weighted Email Network

Corresponding to real-world organizational structure, email networks have a natural property: community structure. In this paper, we propose an algorithm to detect the community structure in weighted email networks by deleting all the boundaries. In order to measure how much an edge could be a boundary between two communities, a composite index named mediumness is defined, which is derived from betweenness centrality. After the graph becomes unconnected via removing the boundary edges, two inspecting criteria are employed to identify the qualifications of sub-graphs for being communities. We test the algorithm on a large computer-generated network which is constructed by randomizing rules. The results show that it can detect all the potential communities in this email network.