Complex Network Community Detection Based On Particle Competition

Community research on complex networks has become increasingly prominent in reflecting the internal structure of network organizations,target group expansion and disintegration,as well as analysis,prediction,and guidance for information propagation.However,the current community research in complex networks has a series of problems that need to be further improved: 1.Some community detection algorithms are not accurate enough due to their simple logic and few parameters.2.The number of communities in some community detection algorithms needs to be pre-designated,which leads to unobjective number of communities and the detection results.3.Current community detection algorithms for dynamic evolution use methods such as time slicing or local evolution,and it is difficult to detect communities in an evolved network in a time series from a global perspective.These problems further affect the research on accurate community detection,network reconstruction and segmentation,and the evolution of network.In addition,the non-linearity,intelligence,dynamics,and emergence of the particle competition mechanism can enhance the objectivity and timing characteristics of community detection and solve the above problems.Therefore,this thesis has carried out the detection and evolution of complex network communities based on particle competition mechanism with the following three aspects:1.A parameter adaptive community detection model has been proposed.Based on the classic community detection algorithm of stochastic competition learning model,using methods such as node similarity,particle swarm optimization,and domination ability adaptation,the parameters of proposed model can be self-adapted and the accuracy of that model has also been improved compered with the original model.2.A method for calculating the number of communities in dynamic evolving networks has been proposed.In this thesis,the calculation of the number of communities in dynamic networks has been realized by constructing a particle competition model and other methods based on the particle competition mechanism.By constructing the calculation model of the number of dynamic network communities,the changes in the number of communities in the network can be obtained intuitively,and the evolution trend of the network has also been analyzed.3.A dynamic community detection model based on particle competition has been proposed.The jump mechanism has been introduced to construct the dynamic particle propagation model,which realizes an effective dynamic community detection function.Compared with the original particle competition models and other dynamic community detection models,the accuracy of the proposed model has been improved.And it has time smoothness which can reflect the dynamic characteristics of the evolution of the community.