Research Of Community Detection In Complex Networks Based On Weighted Label Propagation

In the complex network, nodes are divided into groups, the nodes in same group are closely connected, and the nodes in different groups are sparsely connected, this feature of complex network is called community structure. In the era of large data, discovering community structure accurately in complex network, especially in large-scale complex network, is one of the important problems in the area of community detection.In this paper, we study complex network community detection algorithm based on non-negative symmetric matrix factorization and label propagation method and propose the weighted label propogation algorithm(MFWLP and OMFWLP) to detect the community structure. A node’s influence measurement is proposed and is used to address the problem of random selection of label in label propagation algorithm.A node’s influence measurement includes the nodes similarity measurement and the node importance measurement. The similarity between two nodes is calculated by the latent factor of the nodes which can be obtained by the non-negative symmetric matrix factorization. Node importance is measured by the number of neighbor nodes associated with the node. The node’s influence measurement is used as the weight of label in MFWLP algorithm which can detect the non-overlapping communities. On the basis of MFWLP algorithm,we introduce the node’s label library and propose overlapping weighted label propagation community detection algorithm(OMFWLP).We apply the Map-Reduce model in Hadoop distributed platform to implement the two weighted label propagation community detection algorithms.Experiments were carried out on some real social network data sets and synthetic network data sets. The experiments show that our algorithms are very effective and robust and can detecte the community structures in large-scale networks.