GCN recommendation model based on attention mechanism and user attributes
To further improve the recommended precision of graph convolutional network(GCN)and the convergence speed of the model,a GCN recommendation model based on attention mechanism and user attributes is proposed.The model uses a lightweight GCN to learn the high-order association information of users and items.Then,weighted summation is embedded to different neighborhood features by attention mechanism to get user and item potential feature vectors,and user attribute feature vectors extracted by multi-layer perceptron machine is fused into user potential feature vectors.Finally,the inner product of user and item potential feature vectors is recommended as a prediction result.Experimental verification on the Movielens—1M dataset show that the proposed model is prior to the baseline model.
NETL
NSTL
万方数据
