FRINGE: A new approach to the detection of overlapping communities in graphs

Abstract

Currently, there is a growing interest in identifying communities in social networks. Although there are many algorithms that suitably resolve this problem, they do not properly find overlaps among communities. This paper describes a new approach to the detection of overlapping communities based on the ideas of friendship and leadership, using a new centrality measure, called extended degree. We describe the algorithm in detail and discuss its results in comparison to CFinder, a well-known algorithm for finding overlapping communities. These results show that our proposal behaves well in networks with a clear leadership relationship, in addition it not only returns the overlapping communities detected but specifies their leaders as well. © 2011 Springer-Verlag.