Research On Strict Community Detection Algorithm Of Weighted Complex Networks

Complexity science is a new stage of development of system science, and complex network is one of the important research methods in the study of complex systems.Complex network is a network of some or all of the properties of self organization, self similarity, attractor, small world and scale-free. Studies have shown that community structure exists commonly in the complex networks and studying community (structure) can help us know more about the integral and local character of networks.There are many researches on community structure in unweighted networks and the study has made some substantial progress, but little study of weighted networks community structure has been made, and existing research is not enough in-depth. On the basis of summarizing and analyzing the existing community structure research, this paper studies community structure in weighted networks. The research includes the definition of community structure in weighted networks, weighted artificial scale-free network model with known community structure discovery algorithm about weighted networks.Through a large number of theoretical analysis and experimental verification, this paper obtains the following results:1.put forward an improved strong/weak community structure definition of weighted networksThe definition considerate node degree and average power, which makes the nodes in community meeting the definition has a relatively large number of edges and big edges within the community. Through the statistical analysis of the classical real network, the experiment demonstrates the rationality of the definition.2. put forward the way to build weighted scale-free networks consisting our defined community structure (CWBA model)The method use a unweighted network with a small number of nodes as the initial network, then generate unweighted scale-free network with initial nodes community ownership, next generate weighted scale-free network with initial nodes community ownership, at last adjust the initial nodes community ownership to meet our community definition. Through the above steps, CWBA model is built. In the experiment, two examples are generated according to the CWBA model generation method, and the validity of the method is verified by the statistical analysis of the two examples.3. put forward a community detection algorithm(NPEND algorithm)The algorithm generate and use prior information set, put forward the edge association subordinate degree and the node community contribution two indexes, on the basis of the process of improving the traditional aggregation algorithm, the algorithm results in line with the community in complex networks. Algorithm is verified on the real network and artificial neural networks, the experimental results show the effectiveness of NPEND algorithm.The comparison of the results with other community detection algorithms on the same data set proves the superiority of NPEND algorithm.