Research On Multi-granular Community Detection Methods

Community structure is an important structural property of complex networks. Community detection is an important topic of network analysis. The community structure refers that edges are closely connect inside communities, and relatively loose between communities. Some meaningful entities are usually represented as communities in complex networks. Community detection has the vital significance understanding performance attributes and revealing the internal structure of the complex networks. In recent years, researchers has made some achievements in community detection, and has proposed many community detection methods. In community detection, mainly researching and analyzing community structure of networks in a certain granularity respect. However, many levels of community structure have influence on the network functionality in real networks. It is not easy to analysis the community in single granularity. Therefore community need from the perspective of multi-granularity and multi-level to detect it.In this dissertation, we mainly research the relative issues about community detection in complex networks. According to the information of network topological structure, from the single granularity which based on the neighborhood searching and under multi-granularity that analysis the changes of thickness of network granularity to discover communities. For community detection methods of complex networks, the work is as follows:First, we summarize the research status of community detection algorithm, and briefly analysis the range of application of the various algorithms and their advantages and disadvantages.Second, we propose a community detection algorithm based on the neighbor node under single granularity for community detection method in complex networks. Selecting the node with the largest degree in the network as the community core to search. Then add the neighbor nodes that satisfies certain quantitative criteria to the community, and repeat the above steps to the remaining nodes until no node satisfies the conditions. Finally using neighbor voting method to decide the community for the nodes that not satisfying the conditions, eventually forming a plurality of disjoint communities.Third, we extend studied the community detection algorithms under single granularity in complex networks, and propose the multi-granularity community detection method. Under the single granularity, using effective community detection algorithm to get the information of community structure of the entire network. Then observe the mining communities of network through the change of granularity of network. The known communities to constitute a new network topology structure, continue to repeat mining community. Observe communities under different granularity through the size of granularity until it conforms to the requirements of the analysis.Fourth, we study the community dividing methods, and comparing them with our algorithm with the existing classic algorithms, then verifying our algorithm is feasible and effective on computational efficiency and more accuracy to discover community in complex networks.