基于矩阵分解的推荐算法研究综述  

A review of recommendation algorithms based on matrix factorization

作　　者：程龙 李涵[1] CHENG Long;LI Han(School of Applied Science,Beijing Information Science&Technology University,Beijing 100192,China)

机构地区：[1]北京信息科技大学理学院,北京100192

出　　处：《北京信息科技大学学报（自然科学版）》2021年第2期38-45,51,共9页Journal of Beijing Information Science and Technology University

基　　金：国家自然科学基金资助项目(9011623901);北京信息科技大学教学改革项目(2020JGYB41)。

摘　　要：大数据时代下,现有的推荐系统面临着准确性、数据稀疏性以及冷启动问题的挑战。矩阵分解是解决数据稀疏性的有效方法,近年来,基于矩阵分解的推荐算法备受关注。以矩阵分解为基础,分别概述了基于传统矩阵分解的推荐算法、基于社会化的推荐算法和基于深度学习的推荐算法,分析各类算法之间的优势与不足,对算法的实际应用场景进行总结,最后给出未来研究方向的展望,为后续相关研究提供有效参考。In the era of big data,the existing recommendation systems face challenges of recommendation accuracy,data sparsity,and cold start.Matrix factorization is an effective method to solve data sparsity.In recent years,recommendation algorithms based on matrix factorization have attracted much attention.Based on the matrix factorization method,various recommendation algorithms based on traditional matrix factorization,socialization,and deep learning are summarized.By analyzing the advantages and disadvantages of these algorithms,as well as the actual application scenarios of the algorithm,the future research direction is suggested to provide an effective reference for the follow-up related research.

关 键 词：推荐算法 矩阵分解 社会化推荐 深度学习 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]