Link prediction based on the naive Bayes with high-order clustering structure of node

Link prediction, a significant branch of complex networks, has attracted the attention of a growing number of scholars. It is an important tool in data mining. It is used to predict possible future links in the network or links that have not been observed yet. It can also be used to identify spurious links. The Local Naïve Bayes Model accurately distinguishes the contribution of different common neighbor nodes to the formation of the target link, but it only considers the contribution of common neighbors. A large number of networks have higher-order characteristics, and higher-order structures capture as much information about the network. In the work, we proposed a novel method of link prediction based on Naive Bayes with High-Order clustering structure (NBHO) of node. NBHO not only overcomes the shortcomings of the co-neighbor (common neighbor) similarity index that each co-neighbor of two nodes contributes equally to the likelihood of the connection, but also makes use of a higher order clustering structure. High-order structure plays an important role in the evolution of the network. Compared with the traditional method, this framework can provide more accurate predictions. Obviously there is a conclusion that the higher order structure significantly improves the accuracy of the predictions.