Community Detection Algorithm Based on Node Stability and Neighbor Similarity

Abstract

With the increase of the scale of complex network,community structure becomes more complex.The relationship between nodes and communities become more diversified.It is expected to improve community detection algorithm performance by effectively measuring the community structure and dealing with the nodes with different certainty of community belonging.This paper proposes a community detection algorithm based on node stability and neighbor similarity.Firstly,label entropy of nodes is defined to measure node stability and the nodes with low label entropy are selected as stable node sets.Then the neighbor simila-rity is defined according to the label of node neighbor and the community belonging consistency of nodes and their neighbors is measured.The initial network is constructed by using the node with the highest neighbor similarity between the stable node and its neighbor,and the initial community detection results with high reliability are obtained by running label propagation algorithm on the subnetwork.The unclustered nodes are allocated to the community of the node with the highest Katz similarity.The final result of community detection is obtained by merging small-scale communities.Compared with LPA,BGLL,Walktrap,Infomap,LPA-S and other classical algorithms,experimental results show that the NSNSA algorithm performs well in modularity and NMI.

Keywords

• complex network|community structure|label entropy|node stability|neighbor similarity