| 摘要: |
| 提出了一种基于知识图谱（KG）的用户多偏好（MPKG）推荐系统,从用户关系级、实体级和细粒度高阶用户三种不同的视角建模用户的偏好. 首先,将KG中关系向量组合,构建关系级意图,并通过独立性将不同意图之间的差异最大化,由关系级意图来指导学习关系级偏好；然后,根据用户交互实体的频率构建实体偏好图（EPG）,并学习用户的实体级偏好；接着,分别使用关系级意图和实体级偏好来指导模型学习用户的表示；此外,还直接从KG中构建关系实体信息流,用于用户的表示,挖掘用户的高阶细粒度偏好. 在两个基准数据集上进行实验,实验结果验证了该方法的有效性和可行性. |
| 关键词: 推荐算法 深度学习 知识图谱(KG) 图神经网络(GNN) |
| DOI：10.3969/J.ISSN.1000-5137.2024.02.008 |
| 分类号:TP391 |
| 基金项目: |
|
| A user multi-preference recommendation system based on knowledge graph |
|
LI Jingjing1, ZHAO Qin1,2
|
|
1.College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 201418, China;2.Shanghai Engineering Research Center of Intelligent Education and Bigdata, Shanghai Normal University, Shanghai 200234, China
|
| Abstract: |
| A user multi-preference recommendation system based on knowledge graph （KG） was proposed in this paper, by which user preferences from three different perspectives： user relationship level, entity level, and fine-grained high-order user were modeled. Firstly, the relationship vectors in KG were combined to construct relationship level intentions, and the differences between different intentions were maximized through independence. Relationship level intentions were guided to learn relationship level preferences. Secondly, an entity preference graph（EPG） based on the frequency of user interaction with entities was constructed, and the user’s entity level preferences were learned. Thirdly, relationship level intent and entity level preference were performed to guide the learning of user representations, respectively. In addition, the relationship entity information flow was directly constructed from KG for user representation and mining of high-order fine-grained preferences. Experiments were conducted on two benchmark datasets, which verified the effectiveness and feasibility of this method. |
| Key words: recommendation algorithm deep learning knowledge graph (KG) graph neural network(GNN) |
