Research On Complex Network Community Partition Algorithm Dased On Common Neighbor Node

As a theoretical model for studying complex systems in the real world,complex networks can abstract and summarize the structural characteristics of complex systems,and present the real world networks in an intuitive and concrete form.The community structure of complex network is a cluster with tight internal connection and sparse external connection.The division of community structure in complex networks can reveal the structural characteristics and relevant information existing in the network and discover the intrinsic properties of the network,which has important practical application value and has been widely used in the fields of personalized recommendation,knowledge map analysis and public opinion analysis.Based on common neighbor nodes,this paper studies the community division algorithm for complex networks,and the main work is as follows:(1)In order to solve the problems of high time complexity and low community classification accuracy in the classification of non-overlapping communities by traditional hierarchical clustering algorithm,a community classification algorithm based on similarity of common neighbors was proposed.Considering the influence of co-neighborhood nodes on the similarity between nodes,the algorithm proposed a novel co-neighborhood node similarity model,which constructed a star neighborhood network and measured the similarity between nodes by using the similarity between star neighborhood networks.Moreover,by judging whether the star neighborhood network contains the same nodes,if not,no similarity calculation is performed,so as to reduce the complexity.According to the similarity and local influence of nodes,the initial clustering was carried out,and then the communities obtained from the initial clustering were combined for the purpose of modularity optimization,so as to obtain better community division results.(2)In order to reflect the characteristic structure of the real world network and conform to the actual network situation,a weighted network overlapping community partition algorithm based on the degree of node dependency was proposed.The algorithm constructs a weighted network model,converts the unauthorized network into a weighted network,and reflects the connection strength between nodes through weights.In view of the unreasonable selection of extension nodes in the current local extension algorithm,the core community concept is defined in the transformed weighted network according to the node weight and network topology.The node dependency degree function was defined,and the core community extension was carried out by calculating the dependency degree value between the node and the core community and comparing it with the dependency threshold.According to the idea of extension module degree optimization,the overlapping community division was completed by adjusting the dependency threshold until the optimal community structure was obtained.In synthetic network data sets and real world network data set,this article proposed algorithm and other algorithms using mutual information and module standardization degree as evaluation index,the experimental data show that this proposed algorithm has obvious advantages,can accurately and effectively find the network community structure,divided into the community structure of high quality.