Research And Implementation Of Community Detection Algorithms In Social Network

With the development of mobile Internet technology and Web2.0 technologies,and the rapid spread of mobile intelligent terminal,increasing the participation of people in social networks,further social networking become essential life tool to most people.With so many persons in the social networks makes the network structure complex,which contains a large number of internal information worth to dig.Mining social networks can help us analyze community structure and function of the network topology,and can help us understand the nature of the network then can control and predict social networks.Therefore,the study of large-scale network of community discovery algorithm with clustering high accuracy and low time complexity become a hot topic of scientific research personnel at domestic or foreign.The main work of this paper includes the following points:(1)Elaborating social networking and related concepts and features,and Modeling the mathematical expression of model abstraction of social networking.Reviewing and classifying the traditional community detection algorithms,describing the performance of advantages and disadvantages of each method detail,checking out the latest community discovery algorithm,then we propose two new community detection algorithm.(2)A community detection algorithm based on Genetic Algorithm and Harmony Algorithm is proposed.To the shortcomings of high time complexity,low efficiency and fall into local optimum of the community discovery algorithm based on genetic algorithm.Using labels propagation algorithm to generating initial harmony library,which with better diversity and precision.While using the strategies of dual cross and single point mutation of Genetic Algorithms to generate new solutions,which expanding range of search to improve the clustering precision.Experimental results show that the accuracy of GHS clustering algorithm is better than FN,GN,LPA and K-means spectral algorithm.(3)An automatic community detection algorithm based on spectral clustering by determining the center of FCM algorithm(FCMASC)is proposed.To the disadvantage of K-means spectral clustering,can't automatically determine the number of community and the low clustering accuracy of K-means,an automatic spectral clustering community detection algorithm based eigengap and FCM called FCMASC is presented.This algorithm determines the number of community based on the max eigengap of eigervalues,determines the initial clustering centers of FCM based on the linear correlation of eigenvectors matrix,uses the FCM algorithm clusters the eigenvectors matrix.Experiment tests show that FCMASC improves the clustering accuracy of K-means spectral clustering.