Research on Personalized Recommendation of Link Prediction Based on Multi Graph Attention Network

Graph attention network (GAT) can simultaneously consider the correlation and correlation differences between nodes and their neighbors and can effectively improve the accurate extraction efficiency of node feature vectors. In order to improve the accuracy of personalized recommendation, a link prediction personalized recommendation method based on multi graph attention network is proposed in this paper (MGAT-LPPR). This method uses multi graph embedding, updates and initializes the feature vector, carries out feature learning through the graph attention network, and the output features are used for prediction and scoring, so as to realize link prediction, and finally carry out personalized recommendation. Multi graph embedding can express the high-order information of the model and integrate the adjacent information. The double-layer attention set in the graph attention network. The mechanism obtains the potential components of user (item) characteristics, aggregates these components, and automatically learns the importance of different components. The method used in this paper adopts RMSE and Mae evaluation indicators, and compares with some mainstream algorithms. The experimental results on three real data sets show that the proposed method can achieve better personalized recommendation effect.