计算机工程与应用 ›› 2023, Vol. 59 ›› Issue (21): 102-111.DOI: 10.3778/j.issn.1002-8331.2302-0226

• 模式识别与人工智能 • 上一篇    下一篇

基于距离策略的知识图谱图卷积网络推荐算法

邢长征,刘义海,郭亚兰,郭家隆   

  1. 辽宁工程技术大学 电子与信息工程学院,辽宁 葫芦岛 125105
  • 出版日期:2023-11-01 发布日期:2023-11-01

Knowledge Graph Convolutional Network Recommendation Algorithm Based on Distance Strategy

XING Changzheng, LIU Yihai, GUO Yalan, GUO Jialong   

  1. School of Electronics and Information Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China
  • Online:2023-11-01 Published:2023-11-01

摘要: 图卷积神经网络在处理知识图谱时存在模型训练开销大、元路径设计缺少普适性的问题,针对这类问题,提出了一种基于距离策略的知识图谱图卷积网络推荐算法。通过在知识图谱中划分中心节点和辅助信息节点的方式对知识图谱进行重构,优化生成的邻接矩阵,然后在图卷积算法的基础上,将用户和实体之间的关系建立标准化评分,根据辅助信息距中心节点之间的距离设计了距离策略,通过距离-影响力函数完成对知识图谱中信息特征的提取,最后将知识图谱特征作为辅助完成推荐任务。实验选取三个数据集与其他先进模型进行对比实验,发现该模型在训练时间和推荐效果上均处于优势地位,具有一定的普适性。

关键词: 推荐系统, 知识图谱, 图卷积网络, 距离策略, 知识表示学习, 效率推荐

Abstract: Graph convolutional neural network has the problems of high model training cost and lack of universality in meta-path design when dealing with knowledge graphs. Aiming at these problems, a knowledge graph convolutional network recommendation algorithm based on distance strategy is proposed. The knowledge graph is reconstructed by dividing the central node and auxiliary information node, by optimizing the generated adjacency matrix, and then on the basis of the graph convolutional algorithm, the relationship between users and entities is standardized and scored. A distance strategy is designed according to the distance between the auxiliary information and the central node, and the feature extraction of information in the knowledge graph is completed according to the distance-influence function, Finally, the knowledge graph feature is used as an assistant to complete the recommendation task. The experiment selects three data sets and compares them with other advanced models. It is found that the model is in a dominant position in terms of training time and recommendation effect, and has certain universality.

Key words: recommendation system, knowledge graph, graph convolutional network, distance strategy, knowledge representation learning, efficiency recommendation