Mining Dynamic Communities based on a Novel Link-Clustering Algorithm

Discovering communities in time-varying social networks is one of the highly challenging area of research and researchers are welcome to propose new models for this domain. The issue is more problematic when overlapping structure of communities is going to be considered. In this research, we present a new online and incremental community detection algorithm called link-clustering which uses link-based clustering paradigm intertwined with a novel representative-based algorithm to handle these issues. The algorithm works in both weighted and binary networks and intrinsically allows for overlapping communities. Comparison with the state of art evolutionary algorithms and link-based clustering shows the accuracy of this method in detecting communities over times and motivates the extended research in link-based clustering paradigm for dynamic overlapping community detection purpose.