Research Of Community Detection Algorithm And Community Relation Evolution Model Based On Complex Network

Complex network is a kind of system structure which widely exists in human society and nature.It can be used to capture the complex network system and describe the evolution,evolution mechanism and the dynamic behavior of complex network.And a large amount of complex network that looks different can be dealt with through a variety of universal algorithms.On the one hand,it helps to evaluate and analyze performance of the whole network by learning complex network community detection algorithms.On the other hand,it can also provide a theoretical basis for the design of network protocol,and provide the basis for improving network security and resource optimization.First of all,this paper analyzes and compares several community detection algorithms that have been presented in complex network including image segmentation method,clustering algorithm,splitting algorithm,spectral method,dynamic algorithm,statistical inference algorithm based on module and method based on degree and so on.Secondly,the process and performance of information entropy of self-organized classification algorithm is analyzed and completed.Aiming at the deficiencies of the algorithm,an adaptive algorithm based on getting an average amount of information for members of the community is proposed and completed.And not only the algorithm is designed and realized,but also the performance of the algorithm is analyzed.What's more,the solid growth models in complex network are established including the single node growth model,block growth model and growth model based on the difficulties of communication.And the models are theoretically analyzed and simulated to verify whether these growth models have characteristics of high robustness,high clustering and low average path.At last,the internal balance mechanism of the network of relationships is studied.How to increase,delete,update and change the relationships among the internal network nodes is analyzed in order to reflect the relationships between entities more accurately.Experiments show that the final model can generate scale-free network that has characteristics of relationship strength,high aggregation and high cluster.