Research On Iterative Clustering Based Community Detection Algorithm And Application

With the explosion of massive data,the data has portrayed a multi-relational and multi-dimension network and provided tremendous opportunities and challenges for the quantitative study of the evolution of the network structure.The research results also help to improve information services and create socio-economic value.There are kinds of sizes of community structures in the network.The study of the community structure in the network is an important way to understand the structure and function of the entire network.Based on the complex networks theory,this thesis builds a local iterative model by using the edge clustering coefficient of advanced extension as an indicator of similarity.On this basis,this algorithm is researched and analyzed from two aspects:real network and synthetic network.At the same time,applying the algorithm to analyze the overlapping degree on the double-layer network can better describe the internal relationship between the double-layer networks,which will provide a good method for the analysis of the double-layer network.The contributions of the thesis are specified as follows:1)The realization and analysis of community detection algorithm based on iterative clustering of advanced extension.Based on the idea that the next-nearest-neighbour node of a node also has a great influence on the result of community clustering,it leads to the idea that the edge clustering coefficient can be advanced extended as the similarity indicator of the node.Therefore,the algorithm is not limited to considering only the directly connected nodes' influence on community detection,but on the basic community detection,it continues to consider the effect of the network's next-nearest neighbor node on clustering;then it follows three kinds of local iterative models: the influence from directly connected nodes,the influence of shared neighbors,and the influence of unique neighbors.Through these three iterative models,the nodes in the network are clustered,similar nodes are slowly clustered together,dissimilar nodes gradually move away from each other,finally form a stable state,and then the community can be easily divided.By applying this algorithm to synthetic network and real network,it can be found that the algorithm has high accuracy and robustness.2)The degree of overlapping on a double-layer network is calculated using this algorithm.The analysis of the overlapping degree on the double-layered network canreveal the coupling degree of the corresponding nodes on different networks,analyze the roles and functions of the nodes in the network,and understand the interdependent relationship between different networks.Therefore,the algorithm is applied to a double-layer network to perform community detection,and the NMI value of each layer of the network is calculated by the obtained division result.Finally,the overlapping degree on the double-layer network is calculated.By applying the algorithm to the neuron network of C.elegans,the network of microblog forwarding,and the protein interaction network for overlapping analysis,it can be found that the algorithm still has high accuracy for detecting the overlapping degree of the double-layer network and reveal the interdependence between different networks.