A kind of community detecting algorithm based on modularized label propagation

Abstract

The discovery of high-quality community is a hot spot in social network analysis and many algorithms have been proposed to discover communities. To find potential community structures, a community discovery algorithm based on label propagation will propagate the label of nodes in social networks, but this method contains uncertainty and randomness, and is very sensitive to the structure of social networks, which causes the final result to be highly unstable and contains a huge number of small and fragile societies. Therefore, a novel called Modular-Label-Propagate-Based algorithm for community discovery is proposed. This algorithm starts with the propagation of the network vertexes, and then binds the nodes with compact structure. Label propagation of the vertexes is executed according to the order of the degree. The thought of greedy method is applied, which works according to the sequence of the neighbor node size of current vertexes, if the module value increases, the sequence will be renewed. Experiments have been conducted on data sets with different characteristics. The experimental results show that modular label propagation algorithm can significantly improve the quality, effect and stability of the found communities, and be close to a linear complexity.