Research On Link Prediction Of Directed Networks Based On Node Importance And Fusion Attributes

Complex networks generally contain two kinds of information: one is the network topology information formed by the interconnection between nodes,and the other is the attribute information of nodes.Accordingly,there are two methods to improve the accuracy of link prediction: one is to represent the network structure more reasonably,and the other is to make full use of the node attribute information.However,the existing topology based link prediction algorithms seldom consider the contribution of the node’s own importance to the similarity,and most algorithms do not consider the node’s attribute information,so there are some limitations in the use of network topology information.Based on this,the improvements made in this paper are as follows.Firstly,in view of the current situation that most of the existing link prediction algorithms focus on the public information of node pairs,but ignore the node’s own information,a directed network link prediction algorithm based on node importance is proposed.This algorithm first calculates the importance value of each node according to the Page Rank idea,then defines the reciprocal link weight by using the local structural characteristics,integrates the importance of the node into the adjacency matrix of the network to obtain the weighted adjacency matrix of the network,then uses the improved bifan prediction algorithm to calculate the similarity of node pairs,and finally extends the idea of this algorithm to the classical algorithm index for link prediction.Secondly,in view of the current situation that the research idea of link prediction algorithm is too single,people pay less attention to network structure attributes and node attributes,and can not fully mine network topology information,a directed network link prediction algorithm based on fusion attributes is proposed.In this algorithm,firstly,the structural attributes based on public neighbors in directed networks are analyzed,and the structural attribute similarity is defined.Then,for the current situation that the global algorithm has high time complexity and can not be applied to large-scale networks,the node attributes are mined by using the local importance based on nodes.Finally,the related concepts and descriptions of fusion attributes are given,and the similarity contribution of node fusion attributes to predicting the impact of unconnected nodes is analyzed.Finally,through experiments on real directed network data sets,the directed network link prediction algorithm based on node importance and the directed network link prediction algorithm based on fusion attributes are compared and analyzed with classical algorithms,so as to verify the effectiveness of the two algorithms in improving the accuracy of link prediction.