Computer Engineering and Applications ›› 2023, Vol. 59 ›› Issue (17): 123-131.DOI: 10.3778/j.issn.1002-8331.2205-0416

• Pattern Recognition and Artificial Intelligence • Previous Articles     Next Articles

Knowledge Graph Attention Network Recommendation Algorithm Combined with User’s Perspective

ZHANG Xiao, LIU Yuan   

  1. 1.School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, China
    2.Jiangsu Provincial Key Laboratory of Media Design and Software Technology, Wuxi, Jiangsu 214122, China
  • Online:2023-09-01 Published:2023-09-01

结合用户视角的知识图注意力网络推荐算法

张潇,刘渊   

  1. 1.江南大学 人工智能与计算机学院,江苏 无锡 214122
    2.江苏省媒体设计与软件技术重点实验室,江苏 无锡 214122

Abstract: At present, most knowledge graph based recommendation algorithm only focus on how to enrich the  representation of the item in recommendation system, but they ignore the collaborative information of the user-item interaction from the user perspective. And they fail to combine the collaborative information with the side information of the knowledge graph at a fine-grained level. Aiming at the above problems, this paper proposes a knowledge graph attention network recommendation algorithm combined with the user’s perspective. On the item side, the model captures the high-order information in the graph by knowledge-aware attention embedding propagation. On the user side, this model devises the user perspective factors to refines the collaborative information in the user-item graph, and aggregates along the path of relational dependence in the knowledge graph. The model further enriches the representation of the user and the item to predict how likely the user would adopt the item. Finally, the experiments on the Last.FM and MovieLens-20M public datasets show that compared with the current mainstream baseline, the model has improved by 13%~22% in the Recall@K indicators and 0.6%~4.5% in the AUC and F1 indicators.

Key words: knowledge graph, graph neural network, recommendation algorithm, attention mechanism

摘要: 目前基于知识图谱的推荐方法大多仅仅关注于如何用知识图谱中的信息丰富项目的嵌入表示,而忽略了从用户视角出发,更细粒度地挖掘用户-项目交互的协同信息,将用户-项目图中的协同信息与知识图谱中的辅助信息充分结合。针对上述问题,提出了一种结合用户视角的知识图注意力网络推荐算法。在项目端,结合知识感知的注意力嵌入传播,捕获图中的高阶信息;在用户端,建模用户视角因子细化用户-项目图中的协同信息,在知识图谱中沿着关系依赖的路径聚合,进一步丰富了用户和项目的表示,进而预测用户对项目的点击概率分数。在Last.FM和MovieLens-20M公开数据集上的实验表明,与目前主流基线相比,该模型在Recall@K指标上提升了13%~22%,在AUC和F1指标上提升了0.6%~4.5%。

关键词: 知识图谱, 图神经网络, 推荐算法, 注意力机制