Abstract: Social recommendation based on graph neural networks learns the embedded relationships between users and items through the information of social graphs and interaction graphs to get the final recommendation results. However, the existing algorithms mainly utilize the static social graph structure, which is unable to mine the potential linking relationship between users, and at the same time do not solve the noise problem in the user-item interaction behavior. Therefore, a self-supervised social recommendation algorithm incorporating residual networks is proposed. Firstly, the algorithm employs a variational hypergraph auto-encoder for link prediction in social networks to obtain a reconstructed social graph, which is used to mine the positive link relationships hidden among users. Secondly, an attention mechanism is utilized to assign different attention coefficients to the original and the reconstructed residual social graphs to obtain a more accurate representation of users. Lastly, to alleviate the problem of noise in the data, an adaptive hypergraph global relation extractor is constructed. Self-supervised signals are created using local embedding information and global embedding information in collaboration with this extractor, which optimizes the local embedding representation and thus mitigates the effect of noise. The algorithm is experimentally compared with baseline models such as NGCF, LightGCN, and MHCN on three datasets, Ciao, Epinions, and Yelp. On the Ciao dataset, Recall@10 improves by 17.1%~48.5%, NDCG@10 improves by 1.4%~37.9%; on the Epinions dataset, Recall@10 improves by 8.3%~56.2%, NDCG@10 improves by 3.7%~29.8%; on the Yelp dataset, Recall@10 improves by 9.1%~53.3%, NDCG@10 improves by 10.7%~66.6%. The experimental results show that the algorithm has good recommendation performance compared to the benchmark model.

Key words: social network, recommendation system, graph convolutional neural network, hypergraph, self-supervised learning

摘要： 基于图神经网络的社交推荐算法，通过社交图和交互图的信息来学习用户和项目的嵌入，得到最终的推荐结果。但是现有算法主要利用静态的社交图结构，无法挖掘用户之间潜在的链接关系，同时也没有解决用户与项目交互行为中的噪声问题。因此，提出了一种融合残差网络的自监督社交推荐算法。首先，算法采用变分超图自编码器对社交网络进行链接预测，得到重构的社交图，以此来挖掘隐藏在用户间的积极链接关系；其次，利用注意力机制为原始社交图和重构后的残差社交图分配不同的注意力系数，得到更加精确的用户表征；最后，为了缓解数据中的噪声问题，构建了自适应的超图全局关系提取器，在该提取器的协作下利用局部嵌入信息和全局嵌入信息创建自监督信号，从而优化局部的嵌入表示，进而缓解噪声影响。该算法在Ciao、Epinions和Yelp三个数据集上与NGCF、LightGCN、MHCN等基线模型进行对比实验，在Ciao数据集上Recall@10提升了17.1%~48.5% ,NDCG@10提升了1.4%~37.9%；在Epinions数据集上Recall@10提升了8.3%~56.2% ,NDCG@10提升了3.7%~29.8%；在Yelp数据集上Recall@10提升了9.1%~53.3% ,NDCG@10提升了10.7%~66.6%。实验结果表明该算法相较于基准模型有良好的推荐性能。

关键词: 社交网络, 推荐系统, 图卷积神经网络, 超图, 自监督学习