Improved Overlapping Community Detection Algorithm Based on SLPA
SLPA(Speaker-Listener Label Propagation Algorithm)has linear time complexity and excellent detection effect on overlapping community detection tasks,but as a random algorithm,its repeated random selection strategy limits the accuracy of the algorithm and the results of the algorithm are unstable.In addition,when the selected threshold is low,a large number of small communities and overlapping nodes are easy to appear.Aiming at the above problems,an improved algorithm with higher accuracy and better stability is proposed.In the initialization stage of the algorithm,the ascending order of node importance calculated by node local structure entropy(LE)is used as the node update sequence.In the stage of label propagation,the resource allocation(RA)is used as the basis for further selection of nodes to guide the direction of label propagation.In the post-processing stage,pairwise comparison to the selected community set is added to remove the nested communities.The proposed algorithm is verified on real networks and artificial networks,and compared with five classical algorithms by using overlapping normalized mutual information(NMIov)and Extended Modularity(EQ).Experiments show that the improved algorithm has advantages in accuracy compared with the classical algorithm,and has good robustness in both real networks and artificial networks.Compared with the original algorithm,the results of the improved algorithm are more concentrated and the stability of the algorithm is improved.
complex networkoverlapping community detectionlabel propagation algorithmlocal structure entropySLPA
国家科技期刊平台
NSTL
万方数据
