Heterogeneous information-based self-supervised graph learning for recommendation

In recent years, graph neural networks (GNNs) have become a powerful tool for graph representation in recommendation systems. However, in real recommendation systems that often involve heterogeneous graph data, existing recommendation models find it difficult to learn accurate embedding representations from heterogeneous graphs, leading to the degradation of recommendation model performance. To address these problems, this paper proposes a self-supervised graph learning recommendation model (HSGL) based on heterogeneous information. Firstly, a lightweight graph convolution operation is used to complete feature embedding propagation learning in heterogeneous graphs, secondly, a semantic fusion module is designed to map the embeddings of heterogeneous relations into the same feature space, and a heterogeneous graph neural network is combined with self-supervised learning for feature representation enhancement through contrastive learning between different relations, and the final recommendation results are computed based on the fused feature vectors. Our extensive experiments on three real-world datasets show that the model in this paper outperforms various mainstream recommendation models while validating the effectiveness of key parts of the model.