Overlapping community identification approach in online social networks

Abstract

Online social networks have become embedded in our everyday lives so much that we cannot ignore it. One specific area of increased interest in social networks is that of detecting overlapping communities: instead of considering online communities as autonomous islands acting independently, communities are more like sprawling cities bleeding into each other. The assumption that online communities behave more like complex networks creates new challenges, specifically in the area of size and complexity. Algorithms for detecting these overlapping communities need to be fast and accurate. This research proposes method for detecting non-overlapping communities by using a CNM algorithm, which in turn allows us to extrapolate the overlapping networks. In addition, an improved index for closeness centrality is given to classify overlapping nodes. The methods used in this research demonstrate a high classification accuracy in detecting overlapping communities, with a time complexity of O(n2).