A Cultural Resource Recommendation Model Based On Graph Neural Networks

Recommender systems provide an effective solution to the problems of information overload and information insufficiency. The traditional collaborative filtering methods, however, suffer from cold start and sparsity problems in the practical application of recommender systems. Therefore, scholars have introduced auxiliary information to improve the performance of recommender systems. In some existing models, the feature vectors of user-item interaction information are not accurate enough. In this paper, we propose MKR-Bine, a graph neural network based recommendation model, which fully exploits user and item features by extracting user and item feature vectors using a bipartite graph neural network. Moreover, we use the knowledge graph as auxiliary information to improve the accuracy and interpretability of the recommender systems by learning higher-order interaction information between items and knowledge graph entities through cross-compression units. Finally, the performance of our proposed model is tested on the publicly available datasets of recommendation scenarios. The experimental results show that our model achieves significant benefits on real datasets, and the performance improvement is more obvious on sparse datasets.