知识图谱嵌入的高阶协同过滤推荐系统  

High-order Collaborative Filtering Recommendation System Based on Knowledge Graph Embedding

作　　者：徐兵 弋沛玉 王金策 彭舰[1] XU Bing;YI Pei-yu;WANG Jin-ce;PENG Jian(College of Computer Science,Sichuan University,Chengdu 610065,China;Department of Computer and Information Engineering,Shanxi Institute of Energy,Jinzhong,Shanxi 030600,China)

机构地区：[1]四川大学计算机学院,成都610065 [2]山西能源学院计算机与信息工程系,山西晋中030600

出　　处：《计算机科学》2021年第S02期244-250,共7页Computer Science

基　　金：四川省重点研发计划(2020YFG0089);四川大学-自贡市校地科技合作专向基金(2018CDZG-15);山西省青年科技研究基金项目(201801D221176)。

摘　　要：针对推荐系统存在的数据稀疏问题,传统的协同过滤方法无法捕捉辅助信息之间的相关性,从而降低了推荐的准确度,文中提出KGE-CF模型,引入了知识图谱作为辅助信息,利用知识图谱中多源结构性的数据来缓解数据稀疏问题。KGE-CF结合多层感知机捕获高阶非线性特征的能力,能够学习出用户与项目更深层次的交互信息,从而提升推荐质量。首先,KGE-CF模型将用户的历史交互项目与知识图谱中的实体进行映射,并且利用知识图谱的翻译模型进行训练,得到实体嵌入向量与关系向量,并依据“兴趣迁移”思想进一步学习出更为丰富的用户向量;然后,模型将学习得到的用户向量与项目向量拼接,作为多层感知机的输入,捕捉用户与项目之间的高阶特征信息;最后,通过一个sigmoid函数得到用户对候选项目的偏好程度。通过在真实数据集上的实验,证明了提出的KGE-CF模型在点击率预测和top-k两种推荐场景下均优于其他方法。For the problem of data sparsity in recommendation systems,traditional collaborative filtering methods fail to capture the correlations between auxiliary information,which decreases the accuracy of recommendation.In this paper,we propose the KGE-CF model by introducing the knowledge graph as auxiliary information,and utilizing the multi-source structured data in knowledge graph to effectively alleviate data sparsity.KGE-CF integrates multi-layer perceptron to capture the high-order nonlinear features to learn deeper interaction information between users and items,which can improve the quality of recommendation.Specifically,we first map the user’s historical interaction items with the corresponding entities in the knowledge graph,and use the translation model of the knowledge graph for training,through that we can get the entity embedding vector and the relation embedding vector,moreover the model can learn a richer user vector by interest disseminating.Then,after concatenating the obtained user vector and the item vector,we send it into the multi-layer perceptron to capture the high-order feature information between the user and the item.Finally,a sigmoid function is used to obtain the user’s preference probability for candidate items.The experimental results on the real-world datasets prove that the proposed KGE-CF model in this paper has achieved the best recommendation performance than state-of-art methods.

关 键 词：知识图谱 多层感知机 推荐系统 协同过滤 数据稀疏性 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]