Link Prediction Algorithms Based On The Contribution Of Networks Nodes

With the rapid development of information technology,a mass of complex systems emerged in life,and network scientific research has developed rapidly.Link prediction problem is one of the important research branches,which is used to predict the possibility of future links between unbounded nodes in the network.The research of link prediction can not only help to understand the evolution mechanism of complex network,but also have a wide range of practical application value in different fields such as commodity recommendation,prediction of interaction between proteins and information push.The structure characteristics of network topology are easy to acquire and more reliable.Most of the existing predictors based on network structure similarity measure the similarity between nodes according to the information of topology structure,but do not consider the difference between each node,that is,the contribution of different nodes to the link is different.Therefore,a new link prediction algorithm based on the contribution degree of nodes is proposed in this paper.Firstly,according to the process of network resource allocation,the definition of node contribution degree is put forward,and the size of node contribution degree is measured by two kinds of node importance indexes: medium centrality and near centrality.The link prediction indexes based on node contribution degree,namely RBC and RCC,are proposed.The experimental results on many different networks show that the RCC index can obtain a better prediction effect,and also verify the effectiveness of the new index.Secondly,in order to improve the prediction accuracy,the concept of node contribution degree is combined with the local information of the network.On the basis of the link prediction indexes based on local similarity,the information of the contribution degree of the common neighbor node is considered,and six improved predictors are proposed.By analyzing the experimental results on real network,the results show that the prediction accuracy of these new indexes has been improved to some extent.Finally,most prediction methods based on similarity start from the common neighbor of the predicted node,ignoring that the closeness of the connection between the predicted node and its common neighbor will also affect the prediction results.Therefore,in order to further improve the prediction performance of the algorithm,this paper combines the connection tightness between the predicted node and its common neighbor node and the contribution degree of the common neighbor node,and proposes new prediction indexes,namely CRBC and CRCC indexes.Experimental results on multiple networks show that CRBC and CRCC indexes can obtain better prediction results.In this paper,a large number of experiments are carried out on all the proposed indexes,and the results show that these indexes can improve the accuracy of prediction to a certain extent,and it is of great significance to the in-depth study of prediction algorithm.