Abstract: Link prediction is an important research direction of complex networks.The accuracy of current link prediction algorithm is limited due to limited network information available.In order to improve the link prediction performance of complex network,the improved AdaBoost algorithm is used to predict the link.Firstly,according to the complex network samples,the adjacency matrix is established,and the connection relationship between network nodes is constructed.Then the AdaBoost algorithm is used for classification training,and the prediction results are obtained by weight voting.Finally,considering the imbalance of the distribution of positive and negative errors in the prediction of complex network structure,the weight readjustment factor η and its adjustment range are set [η₁,η₂].The weight of multiple weak classifiers in AdaBoost algorithm is dynamically adjusted according to the value to obtain accurate link prediction results.Experiments show that,compared with other common network link prediction algorithms and traditional AdaBoost algorithm,the improved AdaBoost algorithm has obvious advantages in prediction accuracy,and when there are a large number of nodes,the difference of prediction time performance between the improved AdaBoost and other algorithms is small.

Key words: AdaBoost, Adjacency matrix, Complex network, Link prediction, Weight adjustment