Research On Overlapping Community Detection Algorithm Based On SLPA

With human society entering the era of network information,the production and life of people are full of numerous and complicated data.As an important carrier of information dissemination,the study of structure of networks has gradually become an important research topic in the field of network science.The academic circle names it community detection.Discovering the community structure in complex networks has important theoretical significance and realistic value for studying the nature,function and evolution of networks.Overlapping community detection algorithms are more in line with the real network partition than non-overlapping community detection algorithms in the traditional community detection algorithms.Therefore,this paper proposes two improved overlapping community detection algorithms based on the classical label propagation algorithm SLPA.The main contributions of the paper are as follows.(1)The algorithm of DSLPA aims at the randomness of SLPA in the phase of label update and label propagation.The previous stage uses the PageRank algorithm to determine the node label update order.The latter stage combines the improved Jaccard similarity index to make the second decision when the label selection is not unique.In turn,the uncertainty caused by randomness in the original algorithm is improved.(2)The algorithm of MSLPA combines SLPA with the modularity optimization idea.In the initial stage of the algorithm,maximizing the ratio of modularity for rough clustering.At the same time,using the resource allocation index(RA)to select the label in the label propagation phase.When the result is not unique,the method of selecting the highest frequency of occurrence is adopted to determine the updated label,which makes the community detection results are more in line with the real network structure.The algorithm of DSLPA and MSLPA are experimented on five artificial datasets and six real data sets.According to the index normalized mutual information(NMI),extended modularity(EQ)and partition density(PD),both algorithms have significant advantages.At the same time,the algorithm converges faster and the results are stable,which proves that the proposed algorithms hava better implementation and robustness.