Research Of Community Detection In Complex Network Based On Gravitation

With the development of network and increasing the amount of data, The scale and the structure of complex networks have been bigger and more complex,and more and more information included. How to find and mining the useful information from complex networks has become a research emphasis on the network. The emergence of community structer makes the complex networks can be refined and the study of other aspects can be advanced. So it has been got the attention gradually of the researchers, also certain research results have obtained.But previous studies only applied to a single direction without direction cross and the limitation to the research process is existed.In this paper, the model of gravity relationship is introduced into the complex network.Through the quality of network nodes is defined and the gravitational relationship between nodes is established,the gravity can be used to detect community structure. At first,a controllable granularity community detection algorithm based on gravitational search is proposed to discover the relatively independent community in the networks, and different particle sizes of community structure can be obtained. Via the core nodes in the network are found, and community frameworks are build, and nodes are searched by gravity. Then an overlapping community detection based on gravitational measurement is proposed to detect the overlapping community in the networks, and the overlapping nodes are fuzzy devided.The core nodes in the network are found, and the absolutely attracted nodes are attracted from the communities by gravity in the process of multiple iterations, the membership degree from overlapping nodes to the communities are calculated by gravity in the end. Finally, a dynamic evolution model of community based on gravitational reconstruction is proposed to explore the process of community evolution in different time tag. Gravity relationship in each community and among the communities are refactored to adjust the structure of the communities in the networks. the evolution result can be obtained and compared to the static result. The above algorithms this paper propoesd are based on the theory of the same model,and each is interrelated with one another.Experiments in real networks, community structure of complex networks can be more effectively and accurately excavated via the proposed algorithms. Compared with other previous algorithm, the presented algorithms have some advantages as follows: Any prior knowledge and parameter does not required in our algorithms. Different particle sizes of community structure can be got, however only one result can be got for other algorithms. The time complexity of the algorithms proposed in this paper are all nethermore and can be used in the large-scale complex networks.