Link prediction method of Wiki-RfA network based on bidirectional recurrent neural network model

The invention discloses a link prediction method of a Wiki-RfA network based on a bidirectional recurrent neural network model. The algorithm aims to predict existence of a directed network link. The method mainly comprises the following steps of dividing a Wiki data set according to years, extracting a relationship between voters and votes to construct an adjacent matrix, extracting voted evaluation information of the voters, constructing an RNN-based bidirectional LSTM sentiment analysis model, obtaining the weight of each edge according to semantic information, calculating an out-degree score and an in-degree score of each node, calculating an SMA score between the two nodes according to the node scores, calculating the similarity between the PA and the CN, calculating SmaPA and SmaCN, and carrying out multiple experiments to obtain an average value. A structural-based link prediction method usually only considers whether links exist or not, but neglects the directivity between node links. And the link pre