Research On The Application Of Node Importance In Community Division

Some core nodes in complex networks have great influence on other nodes in networks, so finding these core nodes is very important for the study of community structure and node behavior prediction complex networks. In the addition, many technical researches in complex networks are carried in community division, so the research of community division in complex networks is significant. After many years of research, despite the development of many excellent division algorithms, but algorithms for improving the accuracy of community division and reducing the complexity of the algorithm is still facing challenges.In order to solve the problem of low accuracy and low complexity of algorithms, the main focuses of this thesis are as follows;(1) Core nodes in communities must be found accurately before the division of communities, which involves the evaluation of node importance. And the current node importance evaluation algorithm just consider relatively simple factors,which can not accurately find the core nodes in communities. The influence of neighbor nodes on the importance of nodes and their influence on neighbor nodes is taken into consideration, the contribution matrix of node importance is proposed, in which the K-shell value, the average of neighbor nodes and the closeness of nodes are account in the importance contribution matrix as an influencing factor. Then, the evaluation method of node importance is proposed by combining node degree with local center nodes.(2) At present, most hierarchical clustering algorithms are relatively low in accuracy.Multiple core nodes as the initial community to cluster complex networks is used in this thesis. When calculating the similarity between nodes and the initial community, the importance of nodes in the initial community is taken into account, so as to calculate the node and the initial community similarity more accurately, improve the accuracy of the algorithm to divide the community. Finally, the parallel analysis of the algorithm is applied to the distributed platform to improve the efficiency of the algorithm.The experiment of node importance is carried out on two of simple, intuitive, clear structured networks, and compared with a single evaluation indicators. The experiment of community partition is carried out on real network and artificial network, and analyze those results. At the same time, the algorithm is compared with other nodes' importance evaluation algorithm and hierarchical clustering algorithm respectively. And the parallel algorithm is experimented on the large-scale data set and verify the efficiency of the algorithm. The experimental results show that the algorithm of node importance evaluation can accurately and effectively calculate the importance of nodes, and the community division algorithm can quickly and accurately classify the community structure of complex networks. The parallelization algorithm can deal with large-scale complex networks quickly.