Research Of Community Detection Based On Optimized Label Propagation Algorithm

Over the past few decades,with the rapid development of Internet technology,the scale of online network data is exploding.Under the drive of large-scale network data,more and more attention has been paid to the research of complex network.In these researches,community detection,which can discover the structure of communities in the complex network,and can help researchers understand the characteristics and functions of complex network more deeply,and has important theoretical significance and wide application prospect,has gradually become the focus of research in complex network analysis.Community detection is divided into non-overlapping community detection and overlapping community detection.Community structure presented by realistic complex network is usually overlapping,so overlapping community detection is more realistic.At present,owing to the advantages of simple and rapid in community detection,the community detection algorithm based on label propagation is widely used and studied.Among them,COPRA(Community Overlap PRopagation Algorithm),as an extension of LPA(Label Propagation Algorithm),can effectively detect overlapping communities from the complex network,but it also retains the shortcomings of LPA,such as strong randomness,poor robustness,and easily assigning all vertices to a community.In order to improve accuracy and robustness of COPRA,a multi-label propagation algorithm for overlapping community detection was proposed.It used Leader Rank algorithm to quantify the influence and importance of nodes in social network,then extended these nodes to maximal cliques as initial community cores for label propagation according to Leader Rank score,and introduced the importance of nodes into the label initialization and label updating process,which ultimately improved the accuracy of the results of community division.Experiments on LFR benchmark networks and real public datasets show that the proposed algorithm not only improves the stability effectively,but also increases the accuracy for detecting overlapping communities.