The Study Of An Efficient Community Detection Method Using Fuzzy K-core Multi-granularity Decomposition Mechanism

Now,the world is in an age of information explosion,the scale,type and dimension of data is becoming larger and larger.With these massive and complex data,it is particularly important to dig out the hidden effective information in time.Community detection,as one of the important ways to mine the potential information in complex networks,aims to divide each data object into different clusters according to their similarity and closeness,which are called communities.Data objects in the same community have a high degree of similarity and the connections between them,while data objects in different communities have a low degree of similarity and the connections between them.At present,community detection algorithms are widely used in biology,medicine,complex networks and other fields besides social networks.In this thesis,based on the theory of complicated networks,combined with the structural peculiarities of graph networks and the correlation clustering algorithm,an efficient community detection algorithm for large data sets is constructed.The purpose is to maintain its accuracy while significantly reducing the calculation time when calculating the community detection for large data sets.The main content includes the following aspects:1.In order to give an efficient clustering analysis on the overlapping community network,a community detection framework is proposed based on the graph network characteristics,which can transform the overlap community into the non-overlap community.The framework analysises the collection of adjacency points of each node in the network(also called field),then uses the cluster partition algorithm based on connected component to cluster each node neighborhood locally.Branch points consistent with the number of communities in the neighborhood are constructed for each node.Meanwhile,branch points in the other nodes neighborhood are connected to form a new non-overlapping community network branch network,Finally,the community detection is carried out on the branch network and mapped to the original network.Compared with traditional overlapping community clustering algorithm,the experimental results show that the model presented in this thesis has better performance in terms of accuracy and computation time.2.In order to further improve the efficiency of community detection,a community detection algorithm based on fuzzy k-core with multi-granularity mechanism is proposed.In this method,K-core decomposition based on fuzzy membership function is adopted to select the core node set of the whole network,then the subnet composed by the core node set is divided into communities.Finally,the community label of the core subnet is spread to other nodes to complete the global community detection.Experiments on multiple open data sets show that the algorithm maintains good results in the accuracy of community detection and reduces the computation time on large data sets.3.In order to better display the research work,a simple community detection system is designed and implemented.The system is capable of partitioning,analyzing communities and visualizing the finding results in datasets of different sizes.