基于知识图谱信息协同传播的推荐模型  

Recommendation Model Based on Knowledge Graph Information Collaborative Diffusion

作　　者：张海龙 颜金尧[2] ZHANG Hailong;YAN Jinyao(The 15th Research Institute of CETC,Beijing 100083,China;Communication University of China,Beijing 100024,China)

机构地区：[1]中国电子科技集团公司第十五研究所,北京100083 [2]中国传媒大学,北京100024

出　　处：《现代信息科技》2023年第16期94-99,共6页Modern Information Technology

摘　　要：针对当前基于知识图谱的推荐模型没有充分挖掘知识图谱语义结构信息的问题,提出一种融合知识图谱表示学习方法和信息协同传播机制的推荐模型KCOD。KCOD基于经典的知识图谱表示学习模型DistMult与TransR建模并推理实体三元组的语义关系,然后通过交叉计算每一阶历史交互实体向量推理结果与候选物品实体向量推理结果的相似度,进行模型训练及偏好预测。实验结果显示KCOD的性能优于经典对比模型。Aiming at the problem that the current recommendation model based on knowledge graph does not fully mine the semantic structure information of knowledge graph,a recommendation model,KCOD,which combines the representation learning method of knowledge graph and the information collaborative diffusion mechanism,is proposed.KCOD models and infers the semantic relationship of entity triples based on the classical knowledge graph representation learning model DistMult and TransR.Then it performs model training and preference prediction by cross-calculating the similarity between the inference results of historical interaction entity vector and the inference results of candidate item entity vector at each hop.The experimental results show that the performance of KCOD is better than the classical comparison model.

关 键 词：个性化推荐 知识图谱 表示学习 协同传播 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]