融合知识图谱与循环神经网络的推荐模型  被引量:15

Recommendation Model Based on Knowledge Graph and Recurrent Neural Network

在线阅读下载全文

作  者:程淑玉[1] 黄淑桦 印鉴[2] CHENG Shu-yu;HUANG Shu-hua;YIN Jian(Anhui Vocational College of Electronics&Information Technology,Bengbu 233060,China;School of Data and Computer Science,Sun Yat-Sen University,Guangdong Key Laboratory of Big Data Analysis and Processing,Guangzhou 510006)

机构地区:[1]安徽电子信息职业技术学院,安徽蚌埠233060 [2]中山大学数据科学与计算机学院广东省大数据分析与处理重点实验室,广州510006

出  处:《小型微型计算机系统》2020年第8期1670-1675,共6页Journal of Chinese Computer Systems

基  金:国家自然科学基金项目(61472453,U1401256,U1501252,U1611264)资助;安徽省高校自然科学研究项目(KJ2018A0780)资助。

摘  要:为了解决协同过滤推荐所面对的数据稀疏问题,本文提出一个融合知识图谱与循环神经网络的推荐模型.传统的方法是将知识图谱特征学习模块与推荐模块独立开来,这样学习到的实体特征对推荐的帮助不大.本文提出的模型将知识图谱特征学习自动融合到了推荐系统,首先依据"偏好扩散"思想,利用知识图谱中实体的连接获取用户扩散偏好集,其次将用户扩散偏好集作为循环神经网络的输入,融合基于物品的注意力机制进行用户偏好特征表示学习,最后基于用户偏好特征预测用户喜欢某个物品的概率.该模型丰富了用户的偏好特征,学习出对推荐系统更有用的实体特征表示,增强了推荐效果.本文模型在电影和图书推荐上进行了实验,结果表明该模型在点击率预测、Top-k列表推荐等方面比其他相关算法有更好的表现.To address the sparsity problem faced by collaborative filtering,in the paper,we proposed a recommendation model based on know ledge graph and recurrent neural network.Previous approach is to separate the feature learning module of know ledge map from the recommendation module,which learned entity features are not helpful to recommendation.The method proposed in this paper integrates feature learning of know ledge map into recommendation system automatically.Firstly,the user’s expanded preference set is based on the idea of″preference propagation″,obtained by using the relations between entities in the know ledge graph.Secondly,use the user’s expanded preference set as the input of the RNN,introduced object-based attention mechanism to learned the preference feature.Finally,based on user preference features to predict the probability which user like.This method enriches user preference features,learns more useful entity feature representation for recommendation system,and enhances recommendation effect.Applying the method in movie recommendation and book recommendation.Experiment result demonstrated that the proposed method achieved better recommendation performance than state-of-art method.

关 键 词:知识图谱 循环神经网络 推荐系统 偏好扩散 

分 类 号:TP311[自动化与计算机技术—计算机软件与理论]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象