摘要
协同过滤推荐和基于内容的推荐是目前应用于推荐系统中的两种主流手段.传统的协同过滤模型存在着矩阵稀疏问题,基于内容的推荐又不能自动抽取深层特征,且两种推荐手段很难直接融合在一起,无法共同提升推荐系统的性能表现.充分利用了深度学习模型能够深度挖掘内容隐藏信息的特性,将栈式降噪自编码器(SDAE)运用于基于内容的推荐模型中,并将其与基于标签的协同过滤算法结合在一起,提出DLCF(Deep Learning for Collaborative Filtering)算法.经过真实数据集的验证,DLCF算法能够很大程度上克服矩阵稀疏问题,在性能上优于传统推荐算法.
Collaborative filtering （ CF ） and content-based recommendation are the two main approaches widely used in recommenda-tion systems. Conventional collaborative filtering models usually suffer from the sparsity of rating matrices, and contend-based modelslack of the abilities of extracting deep latent features. Normally these two methods cannot be directly integrated to improve the recom-mendation performance. Considering that deep learning algorithms are capable of extracting deep latent features, this paper appliesStack Denoising Auto Encoder （SDAE） to content-based model and proposes DLCF （ Deep Learning for Collaborative Filtering） al-gorithm by combing CF-based model with labels. Experiments on real data sets show that DLCF can greatly overcome the sparsityproblem and significantly improve the efficiency of conventional CF algorithms.
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第1期7-11,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61003031)资助
上海重点科技攻关项目(14511107902)资助
上海市工程中心建设项目(GCZX14014)资助
上海市一流学科建设项目(XTKX2012)资助
上海市数据科学重点实验室开放课题(201609060003)资助
沪江基金研究基地专项(C14001)资助
关键词
推荐系统
协同过滤
深度学习
栈式降噪自编码器
recommendation system
collaborative filtering
deep learning
stack denoising auto encoder
