Research On Social Network Link Prediction Algorithm Based On Node Similarity

Existing link prediction algorithms are mainly based on degree centrality,ignoring the influence of local centrality of the network on node links,so that in the prediction of large network links,the traditional single centrality algorithm cannot achieve better results.In order to improve the accuracy of link prediction for large complex networks,it is necessary to solve the problem of how to integrate the local central features of the network structure in the algorithm.In view of the above problems,the thesis designs a link prediction algorithm that integrates betweenness,aggregation coefficient and degree centrality.The algorithm combines the reciprocal sums of the local centrality indexes(degree centrality,betweenness centrality and aggregation centrality)of the node's common neighbors according to a certain ratio,which is used as the node similarity index.At the same time,the thesis compares the algorithm with other 6 commonly used algorithms based on structural features.In 6 different networks,the data experimental results show that the algorithm scheme has a certain degree of improvement in the accuracy of link prediction compared to the other 6 commonly used algorithms,but it has a higher type of aggregation coefficient centrality and betweenness centrality.In the network(such as Dolphin,Facebook and Wikivote)the increase is more obvious.Considering the complexity of the network in reality,the inherent attractiveness between the nodes of the network also has a certain effect on the links between the nodes.In this paper,a similarity algorithm based on the gravitation of the nodes is designed.The algorithm takes the connection probability between nodes as the quality attribute of the node and the degree of the node as the distance attribute of the node,and constructs the gravitation similarity index of the node based on the law of universal gravitation.For this algorithm,the paper compares the other 9 commonly used structure-based algorithms.Matlab simulation experiments conducted on 6 different networks show that for networks with larger network diameters(such as Fb-pages-food and Fb-pages-tvshow),the prediction accuracy of the algorithm has been significantly improved.