摘要
为解决推荐算法中数据稀疏的问题,利用知识图谱中的语义信息,更加准确地构建用户画像。以DBpedia中电影知识图谱为例,提出将自动编码器的网络结构与基于知识图谱的语义信息结合,赋予隐藏层中的神经元电影主题意义,从用户的观影历史中,得到每个用户对相关主题的偏好程度,完善用户画像的构建,运用协同过滤算法进行推荐。对比实验结果表明,该算法在准确率、召回率等推荐性能指标方面有着良好的表现。
To solve the problem of sparse data in the recommendation algorithm,the semantic information in the knowledge graph was used to build the user profile more accurately.The film knowledge graph in DBpedia was taken as an example.The network structure of the Autoencoder was combined with the semantic information based on the knowledge graph,so as to give the hidden layer of neurons the meaning of movie theme.The preference degree of each user to the related theme was obtained from the records of movie watching,and the construction of user profile was improved.The collaborative filtering algorithm was used to make recommendations.Through comparative experiments,the results show that the proposed algorithm has good performances in the accuracy,recall rate and other recommended performance indicators.
作者
陈涛
刘学军
张伯君
CHEN Tao;LIU Xue-jun;ZHANG Bo-jun(College of Computer Science and Technology,Nanjing Tech University,Nanjing 211816,China;Department of Business and Information Technology,Nanjing Boiler and Pressure Vessel Inspection Institute,Nanjing 210019,China)
出处
《计算机工程与设计》
北大核心
2020年第11期3047-3052,共6页
Computer Engineering and Design
基金
江苏省重点研发计划基金项目(BE2017617)
国家重点研发计划基金项目(2018YFC0808505、2017YFC0805605)。
关键词
推荐系统
用户画像
知识图谱
语义信息
协同过滤算法
recommender systems
user profile
knowledge graph
semantic information
collaborative filtering
