Research Of Social Network Based On Link Prediction

As a branch of link prediction data mining,link prediction provides a bridge for complex networks and information science.An important goal of link prediction is to deal with the reduction and prediction of missing information in information science.Link prediction promotes the development of information science,and has great value in practical application.With the rapid development of social networks,the application of link prediction methods in social networks is becoming more and more popular.In order to adapt to the scale of social networks become large,link prediction methods not only need to ensure that the prediction accuracy is high enough,but also consider the time complexity.The existing link prediction methods can not meet these two requirements at the same time,so it is necessary to improve the link prediction method.In this paper,we study the link prediction in social networks and improve the link prediction method based on local information,the major work and contributions include:1)Research the correlation between the prediction accuracy of the link prediction method and the structure properties of the network based on the structural properties of the network.Due to different definition of similarity,the prediction accuracy in a complex network of different link prediction methods are somewhat different.Through studying the influence of average degree,clustering coefficient and assortativity coefficient on prediction accuracy,next design experiment to calculate prediction accuracy in the actual networks,verify the prediction accuracy of structural attributes and link prediction method of network indeed exist certain correlation.2)On the basis of the correlation research,analyze the different structure characteristics of the social network from other types of networks,according to the "small world" and scale-free characteristics,improve the link prediction method based on localpath,with adding the node degree of node to the contributions of similarity.The experimental results show that the improved method can improve the prediction accuracy without increasing the time complexity.