摘要
协同过滤算法是推荐系统中应用最为广泛和成功的推荐技术,本文针对协同过滤推荐算法中的评分预测问题,对包含正则项的传统BSVD、SVD++模型进行分析改进,详细分析SVD模型的理论方法,加入用户历史行为记录的潜在信息,利用包含用户喜好(如浏览)的隐性特征向量矩阵替换原SVD模型中的用户特征向量矩阵,提出非对称奇异值分解(Asymmetric singular value decomposition,ASVD)模型,并将项目的特征矩阵也进行扩展形成相应的对偶模型,最后将二者的结果进行融合作为最终的预测评分.在Movie Lens数据集上进行实验验证,结果表明基于ASVD的评分预测算法与传统BSVD、SVD++相比,能有效提高推荐系统的预测精度.
Collaborative filtering algorithm is the most widely used and successful recommendation technology in the recommendation system. Aiming at the problem of rating prediction in collaborative filtering recommendation algorithm,an asymmetric singular value decomposition( ASVD) model is proposed,and the theoretical method of eigenvalue decomposition and singular value decomposition is deduced from the perspective of matrix algebra. In addition,the traditional BSVD,SVD ++ model is analyzed. The potential information of the user's historical behavior record is added,and the corresponding dual model is put forward. Finally,the results of the two models were fused as the final predicted rating. The experimental results from MovieLens datasets show that the rating prediction algorithm based on ASVD can improve the prediction accuracy of the recommendation system compared with the traditional BSVD and SVD ++ models.
作者
李春春
李俊
LI Chun-chun;LI Jun(Department of Automation, School of Information Science and Technology, University of Science and Technology of China, Hefei 230026, Chin)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第6期1286-1290,共5页
Journal of Chinese Computer Systems
基金
工业互联网网络架构基础共性和关键技术标准试验验证项目资助
关键词
个性化推荐
协同过滤
SVD
隐语义模型
personalized recommendation
collaborative filtering
SVD
latent factor model