The digital transformation of business and society has led to the growth of networks in almost every field. Finding communities in real world networks has been considered crucial for modern network science. Moreover, the organization of communities into co-occurring disjoint, nested and overlapping structures adds to the complexity of community detection problem. Therefore, methodological rigor is crucial for community detection so as to foster cumulative tradition in data and knowledge engineering. This paper proposes an algorithm for overlapping community detection based on the concepts of rough set theory. Initially, subsets of links are formed by using neighborhood links around each pair of nodes. Subsequently, we iteratively obtain the constrained linkage upper approximation of these subsets. The notion of mutual link reciprocity is used as a merging criterion during the iterations. The proposed algorithm is experimentally evaluated on eight real-world networks. Comparative analysis with state-of-the-art algorithms demonstrates the effectiveness of proposed algorithm.
展开▼
