Research On Link Prediction Algorithm Based On Complex Network Feature Vector Centrality And Clustering Coefficients

Link prediction is one of the important applications to really connect complex networks and computer science,and it deals with the most fundamental problems in information science,so it is very important to study link prediction.How to improve the accuracy of prediction is one of the basic problems in studying link prediction.Most of the current link prediction methods are related to local node metrics based on node similarity,while the nature of neighboring nodes in the network between nodes is often ignored by the metrics of node similarity.To address the above problems,this paper proposes an improved algorithm based on the traditional link prediction based on the node importance,with the following three main research elements.1.link prediction algorithm based on feature vector centrality,which mainly takes into account the feature vector centrality in the importance of complex network nodes to the common neighbor indicator,AA indicator,RA indicator and other node similarity indicators,and uses the classical algorithm effectiveness evaluation method in link prediction-AUC as the improvement algorithm The final experimental results show that the link prediction algorithm based on node centrality is more accurate than AA,RA,CN and other node similarity index algorithms in the link prediction of complex networks,and the improved algorithm improves the prediction effect of the algorithm without increasing the time complexity of the algorithm.The improved algorithm improves the prediction effect of the algorithm without increasing the time complexity of the algorithm,which makes the efficiency of link prediction improved.2.In previous link prediction algorithms,the influence of network density on link prediction is rarely considered,but the influence of network density on the connection between nodes is obvious,because the more dense the network,the greater the degree of a node tends to be,and the clustering coefficient is the performance of network density.Therefore,based on the link prediction algorithm based on eigenvector centrality,this paper also combines the local clustering coefficient formula to improve the efficiency and accuracy of the algorithm,and finally verifies that the results are better than the original algorithm through a series of experiments.The first part is the study of link prediction algorithm combining complex network node similarity index and local clustering coefficient formula.The second part is to integrate the link prediction algorithm based on eigenvector centrality with the improved link prediction formulation by local clustering coefficients mentioned above,taking node centrality and local clustering coefficients into account.The new experiment verifies that the link prediction algorithm combining local clustering coefficients and eigenvector centrality still has better performance on node prediction without increasing the time complexity,and further improves the accuracy of link prediction.3.the study of hybrid information entropy and node importance indicators,the study of complex network link prediction algorithm considering the centrality of feature vectors,combining information entropy with complex network local information indicators,and acting on the basis of common neighbor indicators and resource allocation indicators of link prediction algorithm.Information entropy as a method of evaluating the importance of local information of the network can measure the importance of a node,which can better improve the efficiency of link prediction.The accuracy of the algorithm that mixes information entropy and node importance metrics in generating link prediction between nodes is demonstrated by simulation experiments.