Research On Community Detection Algorithm Based On Complex Network

With the development of science and technology, the research value of complex networks gradually highlight, explore and study the researchers pay more attention to the key nodes in the complex network community discovery, so far, a complex network of community found a large number of research results have been proposed, in which numerous achievements also have some practical methods. Provides convenient conditions for the better research on complex networks. In this paper, by consulting a large number of domestic and foreign literatures under the premise, the complex network community discovery research made a comb and summary, and the complex network theory to study the historical process were summarized, the proposed clustering algorithm pretreatment community.This paper proposes a local clustering algorithm similar to read pretreatment community discovery algorithm, firstly using the community discovery algorithm in constructing similarity can be converted to the basic idea of clustering algorithm, using local similarity index to construct similarity matrix and spectral clustering algorithms, the gap between the standard reference feature preprocessing partition on the current network, and then use the reference index for consideration the global network topology characteristics of the Page Rank algorithm as the core selection, selection of important nodes in each community structure pretreatment in the calculation of non central nodes of each important node of each community structure formed by the value of the contribution that the node to node fitness, select a greater degree of community and then add the corresponding network. The community is divided, finally the fusion algorithm of K-means outstanding ideas, the division results of iterative calculation until Community structure to achieve a stable state. In particular, part of a complex network community detection algorithm for the most important nodes in the network as the core, and community development, when dealing with the core node is not clear when the network partition structure, it is very difficult to guarantee, this method firstly uses the spectral clustering algorithm to test the network community results pretreatment. To overcome the above situation, the initial network structure is guaranteed, at the same time, due to the use of features in the spectral clustering algorithm in the process of calculating the number of initial gap clustering, community structure has certain efficiency to do that because of this, according to the initial division of community as a reasonable number of pre-selected reference value, also avoid K problem the efficiency of means algorithm in the unknown parameters community division K, and on this basis, using the Page Rank node isoparametric fitness The selection of key nodes and the community merging algorithm are further optimized to improve the community structure.