Community Detection Based On Path Similarity

By the study of the social system and information system, people found that the systems can be represented by complex networks. Complex networks have community structure as one of the important structure characteristics in addition to the characteristics of scale-free degree distribution and small world. Network clustering phenomenon in the community structure can make the random nodes clustering in the form of group. This paper proposes two different network community structure detection model,expressed as follows:(1) Community detection of signed networks. In the weighted signed networks,the links between nodes are both positive and negative. According to the rule that different similarity of nodes has different influence on the links between nodes, we proposed path similarity. Through the dynamic evolution of phase node, nodes would form different synchronization clusters, and realize the community detection ultimately. This model is not only suitable for small scale networks, but also has some advantages for the detection of large scale networks.(2) Community detection of positive complex networks. In order to synchronize the phase between the two coupled oscillators and make the phase of two unconnected oscillator asynchronous, the path similarity is proposed in this paper. Different neighbors have different intimate relationship, which means intimate nodes are more likely in a community. Then the similarity of nodes is adopted to describe their intimate. In order to make the two connected oscillators cluster together and two unconnected oscillators get away, similarity based on path is adopted to Kuramoto model. Nodes in the networks will be divided into different synchronization cluster based on this improved model. Karate network and dolphin network are tested to verify the performance of the proposed algorithm. Several simulations results showed that our proposed method is more efficient.