A Rough Connectedness Algorithm for Mining Communities in Complex Networks

Mining communities is essential for modern network analysis so as to understand the dynamic processes taking place in the complex real-world networks. Though community detection is a very active research area, most of the algorithms focus on detecting disjoint community structure. However, real-world complex networks do not necessarily have disjoint community structure. Concurrent overlapping and hierarchical communities are prevalent in real-world networked systems. In this paper, we propose a novel algorithm based on rough sets that is capable of detecting disjoint, overlapping and hierarchically nested communities in networks. The algorithm is initiated by constructing granules of neighborhood nodes and representing them as rough sets. Subsequently, utilizing the concept of constrained connectedness, upper approximation is computed in an iterative manner. We also introduce a new metric based on relative connectedness which is used as the merging criteria for sets during iterations. Experiments conducted on nine real-world networks, including a large word association network and protein-protein interaction network, demonstrate the effectiveness of the proposed algorithm. Moreover, it is observed that the proposed algorithm competes favorably with five relevant methods of community detection.