The Partition Algorithm Research Of Community Structure Based On Node Similarity

Complex networks as an abstraction and description way of complex systems, widely exist in nature and society. The research of complex networks has penetrated into various disciplines. That is to say any complex system can be studied as a complex network, such as computer science, sociology, biology, circuit and system, etc. With the in-depth study of network properties, the researchers found that, these seem to be large and complex system, there are some significant characteristics, such as community structure. Community structure is an important feature structure of complex network, which has been widely used in many ways. Therefore, it is very important to detect community structure in complex networks. In recent years it has been the focus of many research scholars.Complex network community structure partition algorithm is designed to reveal the real network community structures. The researchers proposed a series of efficient algorithms to find the community structure in complex networks, so as to analyze the basic characteristics and common features of community structure in the network. By studying the relationship between nodes, finding similarity and dissimilarity between them, and synthesizing the advantages and disadvantages of the existing algorithm, we propose new community structure partition algorithms. The main content and innovative research of this paper are as follows:(1) A community structure partition algorithm based on node similarity. Firstly, large eigengaps of the normalized modularity matrix are adopted to analyze the multiscale community structure. Secondly, according to the number of common neighbor nodes and shortest path between nodes, the concept of node similarity is defined. Then, based on the node similarity and the definition of the standard community structure, a new community structure partition algorithm is proposed. The modularity Q value function is utilized to evaluate the advantage and disadvantage of community structure division, and the better community scale is selected according to the value of Q. The experimental results show that compared with other algorithms, the proposed algorithm can obtain better network community structure by using less information. Meanwhile, the proposed algorithm is easy to implement.(2) An improved community structure division algorithm based on node dissimilarity. Firstly, the core nodes are sifted and taken as initial core nodes set according to the evaluation norms of degree and average degree. Then it is divided based on the node dissimilarity until finish the community structure division. The experimental results show that, compared with other algorithms, the algorithm is accurate and easy to implement.In summary, the similarity and dissimilarity between nodes from the relationship between nodes are studied in this paper. The larger the similarity between nodes is, the smaller the dissimilarity between nodes is. The more intimate relationship between nodes is, the greater possibility is in a same community. Based on this idea, the partition algorithm of community based on node similarity is proposed.