Density Peaks Based Clustering Algorithm for Overlapping Community Detection

In order to discover overlapping community structure of social networks more effectively, this paper proposes an algorithm of overlapping community detection based on peak density. The algorithm firstly calculates the matrixes of network topology information distance, and then calculates the local density for each node within a given radius. And then cluster centers are those points which have high local density and the longer distance to those points which have higher local density. After obtaining the cluster centers, calculating the probability distribution of nodes within the community, leading to the division of overlapping communities. This algorithm can be used to divide both the overlapping and non-overlapping communities. Furthermore, the algorithm doesn't have to preset the number of communities. Instead, it chooses the number of communities by decision-making map. It shows that the proposed algorithm can separate the network community efficiently from the experimental results of several artificial network data and real data sets.