Multi-attention User Information Based Graph Convolutional Networks for Explainable Recommendation

Dealing with sparsity and cold-start problems in recommendation systems has always been a challenge. We propose a Multi-Attention User Information based graph convolutional networks for explainable Recommendation model (MAUIR), which can aggregate user and item simultaneously through higher-order information. Our model contains two kinds of attention mechanisms-hierarchical attention and inter-level attention. The first is to explore the different contributions of neighboring entities to the central entity, and the second is to capture the influence of the higher-order structure on the central entity. Therefore, these measures are used to better obtain the final representation of the entity, and the model predicts more accurately. In addition, we also add auxiliary information to the dataset to enrich the user representation to make the model more explanatory. Experiments on three data sets show that our model exceeds the baselines.