基于双用户视角与知识图谱注意力网络的推荐模型

Recommendation Model Based on Dual-User Perspective and Knowledge Graph Attention Networks

基于知识图谱的推荐模型可以解决推荐模型中数据稀疏和冷启动问题。为进一步提高推荐模型的可解释性和准确性,提出基于双用户视角与知识图谱注意力网络的推荐模型DKGAT。首先,为获取用户特征,模型从双用户视角分析用户行为并建立用户向量表示。其次,为获取项目特征,模型利用注意力网络在知识图谱中挖掘用户感兴趣的项目属性,自动获取和表示用户的兴趣路径。实验表明,相较于基准模型,该模型在三个公共数据集上的实验结果均有一定提高。

The knowledge graph-based recommendation model can solve the problem of data sparseness and cold start.In order to further improve the interpretability and accuracy of the recommendation model,a recommendation model DKGAT based on dual-user perspective and knowl⁃edge graph attention network was proposed.Firstly,in order to obtain user features,the model analyzes user behavior from dual user per⁃spectives and establishes the user representation.Secondly,in order to obtain the item features,the model uses the attention network to cap⁃ture the item attributes that user is interested in the knowledge graph,and automatically obtains and represents the user's interest path.The experiment results show that our approach outperforms stronger than recommender baselines on three common datasets.