Research Of Community Detection Algorithm Based On Partial Priori Knowledge

With the comming of the era of Data Technology,the value of the data has been more extensive attention in all walks of life.So how to explore some valuable information from the complicated data to guide and improve our work and life has important significance.Community detection is an important research area of complex networks research can find some potential community structures in complex network data,and then found hidden knowledge in the massive network data and potential law under the general phenomenon,and to provide personalise and scientific services and help people make more effective decisions.This article focuses on study of the Label propagation algorithm,by integrate prior knowledge of the process of community detection we proposed a Label propagation algorithm based on local cycle,and validates the proposed algorithm by experiments.The main research work of this paper includes the following two aspects:(1)We proposed a Label propagation algorithm based on local cycle.Firstly,we summarize the algorithms of community detection and analyzed emphatically the Label propagation algorithm and the existing problem.Secondly,according to the prior knowledge of the nodes in the process of community detection,we proposed a Label propagation algorithm based on local cycle.The main idea is in the label propagation process,when there are multiple maximum label,using the shortest local cycle selection policies instead of random selection to suppress the label spread among communities effectively and improve the accuracy of the algorithm.we used a simple example to verify the feasibility of the algorithm from a theoretical point of view.Finally,In order to verify the effectiveness of the improved algorithm,we choose three types of data sets,which is classical real data sets,benchmark data set artificially generated and real microblog data set,and use modularity and NMI as the evaluation criteria to verifiy the improved algorithm proposed in this paper by compared with the Label propagation algorithm.The results showed that the Label propagation algorithm based on local cycle has achieved good effect.(2)Experimental verification.We select the most representative of microblogging real network as experimental data sets,excluding special points by preprocessing and then apply the proposed algorithm to microblogging real network to verify the improved algorithm alse can achieved a good division results in a real network.