Research On Community Detection Algorithms In Complex Networks

There are various kinds of large-scale systems in real life,which can be described by the complex network.In spite of small world and scale-free,the complex network is also significantly characterized by modularity,namely community structure,which reveals some underlying rules and behavior characteristics.In the last ten years,it has been a hot issue that how to effectively explore the community structure in the network.Focusing on this problem,scholars have proposed many pertinent algorithms,in which local expansion method and genetic optimization algorithm are two important classes.By studying a flood of literature,we have exposed such problems in the existing community detection algorithms:(1)The existing algorithm is followed by the problem that the large edge node may be considered the community core node by other extension methods as well as the high time complexity(2)The “premature phenomena” – early convergence in the traditional Genetic Algorithm.In this paper,based on the research of community detection,by analyzing the relationship among nodes;defining correlation coefficient as the standard to measure the relationship among correlation coefficient and the centrality of edges as well as defining the modularity function based on the correlation,two community detection algorithms of the complex network are introduced,as follow:Firstly,a community detection algorithm based on correlation coefficient and gravitation,which chooses two nodes with high correlation as community core nodes;constructs community skeleton by interrelated core nodes,forming the initial community;ranks the left uncategorized edge nodes;calculates the attraction of different communities towards edge nodes;categorizes the node into the community which most attracts it.Secondly,a complex community detection algorithm based on chaos genetic algorithms,which enhances the diversity of group and improve degeneracy problem bounded to the genetic algorithms by introducing the chaos sequence to construct Initializing population.In addition,the search space.is narrowed and the searching capability is improved in that the local network information is brought into full function in the crossover operation.Furthermore,in mutation operation,replace the chosen Genes that need to be mutated with the new genes in the chaotic sequence.Above all,this algorithm can automatically obtain the optimal community number and community partitioning scheme without any prior knowledge.In this thesis,for the sake of validity,a large quantity of real-world web data sets as well as artificial network data sets are adopted to verify these two algorithms,followed by a comparison between the experimental result and the existing algorithms.The results show that both these two algorithms are equipped with stability and efficiency,accurately detecting the community structure of the complex network.