Clustering a network based on the dynamics and interactions of its members

After one of the key objectives for representing real-world problems and systems using networks is for detecting community structures. This is because detecting community structure is crucial for identifying the link between structure and function in complex networks, which is the key for solving many practical applications in various disciplines. The detection of “good” communities has proven to be a challenging task. This is due, mainly, to the fact that most current methods detect communities in independents with sparse connections among them. As a result, most of them do not work well on highly sparse networks. We propose in this paper a system called DCDM that detects disjoint communities and works well on highly sparse networks. It does so by adopting the following procedure: (1) assigning a score to each vertex to reflect its relative importance to the whole network, (2) assigning a score to each link connecting two neighboring vertices to represent the degree of association between them, (3) employing a two-phase strategy for detecting disjoint communities, and (4) enhancing the density of community using a post-processing technique. We evaluated the quality of DCDM by comparing it empirically and experimentally with nine methods. Results showed marked improvement.