Reserrch On Copula MMSB Based Community Detection Algorithm In Weighted Social Network

Social networks are widely used in all fields of the real world.These networks imply underlying structures and interactions and contain rich data and information.Mining the social structures on social networks can help to understand the reasons for network formation and the logical relationship between nodes,Which helps to grasp the topological features and essence of the network.Therefore,the community detection of social networks is a very valuable and realistic research topic.The mixed membership stochastic block model(MMSB)is a community detec-tion algorithm which based on Bayesian network.MMSB was widely studied since it was proposed.However,the MMSB algorithm assumes that the membership indi-cators of each pair of nodes are distributed independently,which is not reasonable in real life,some scholars have proposed the new framework of Copula MMSB,which integrates the Copula function to model the data within the community to complete the community division.However,the current Copula MMSB algorithm ignores the weights of the connected edges between nodes and therefore can not effectively deal with the problem of detection of the directed social network community.In this paper,by studying the existing social network community detection algo-rithm and a weighted social network community detection algorithm based on Copula MMSB is proposed.This algorithm is called weighted Copula MMSB(wCMMSB).The main work is as follows:1)Firstly,the statistical graph model is established and the model building process is described in detail;2)Two kinds of MCMC sampling methods are proposed to estimate the parameters;3)The wCMMSB algorithm is applied to the artificial and real network data.At last,the experimental results show that the proposed wCMMSB algorithm outperforms the classical community discovery algorithms such as CPM,CORPA and SLPA algorithms,both in artificial network datasets and real network datasets.