A Multi-Graph Joint Learning Network with Side Information for Recommender Systems

The application of knowledge graph in recommendation algorithms effectively enhances the interpretability of recommendation results, but it still lacks the mining of deep semantic information. Aiming at the problem that the current knowledge graph based recommendation algorithm is difficult to fully mine the potentially related information between entities, this paper proposes a multi-graph joint knowledge graph recommendation algorithm. This model combines the ideas of metapath2vec and EGES to enhance the embedding representation of nodes by simultaneously learning the original graph composed of the node itself and the auxiliary graph consisting of their side information to improve the recommendation effect. Extensive experiments on public datasets show that, compared with other benchmark algorithms, the proposed approach has a certain improvement in accuracy and robustness, and has a better ability to deal with sparse data.