Research On Link Prediction Based On Topology Information Of Node Attributes

With the advancement of science and technology,more and more complex systems have emerged,and the relevant data generated has increased linearly.These data have also played a role in promoting the research of complex networks.Link prediction is an important research direction of complex networks.The main solution is how to use known data and the interaction relationship between them to predict the data that already exists but has not been observed,the data that may appear in the future,and some Fake data.As the research results of link prediction are widely used in various fields,how to improve the accuracy of link prediction has become a primary issue.The research in this paper is mainly based on the network topology,focusing on the attribute information of the nodes in the network.Starting from the degree,H value and clustering coefficient of nodes,this paper studies how to distinguish the different functions of different nodes in an undirected network.The main research contents of this paper are as follows.First of all,for existing link prediction algorithms,there is no better way to distinguish the different effects of nodes on link prediction.Therefore,the concept of hybrid similarity is proposed,which combines the degree value and H value of the node to further distinguish the different functions of the node.Moreover,among the existing link prediction algorithms,there are relatively few algorithms based on the combination of node attributes and path information.Therefore,the node and path are combined,and a path and node combination prediction algorithm based on hybrid similarity is proposed.Secondly,in link prediction algorithms based on node attributes,one attribute of the node is often used to represent the contribution of the node,but a single attribute cannot fully represent the importance of the node.Therefore,two attributes of nodes,H value and clustering coefficient,are introduced to combine these two indicators.In this way,the importance of nodes is highlighted and the contribution of nodes is reflected.Moreover,most of the link prediction algorithms based on local similarity only consider common neighbor nodes,and the influence of other neighbor nodes is ignored.Therefore,taking the second-order neighbor nodes into consideration,a multi-order neighbor prediction algorithm based on node attributes is proposed.Finally,for the above two algorithms,experiments were performed on different network data sets to verify the effectiveness of the two algorithms.