Research On Complex Networks Time-series Link Prediction Based On Spectral Clustering

Complex network analysis is a popular research field.As an important branch of network research,link prediction can help people not only understand the evolution mechanism of complex network by mining the information of network structure and analyzing the known structure of complex network,but also reveal practical problems in application.This paper summarizes the methods of link prediction,analyzes the factors that affect the results and discusses the optimization methods.Firstly,based on the problem of the incomplete data of input network,a clustering-based fuzzy node identification technique is proposed.The fuzzy nodes existing in the network are clustered and identified under different edge-incomplete degree.By processing the network data,the accuracy of link prediction is indirectly improved.Based on the study of existing link prediction methods,combining nodes inherent clustering tendency and characteristics of scientific cooperative networks,a new link prediction method based on clustering information is proposed for static network analysis(CI).According to the characteristics of the time-series networks,index-weighted and local-path-weighted methods are used to process temporal networks.CI is then extended to time-series link prediction.The above algorithms are tested in several real data sets,and the calculated results are better than the traditional methods.