基于用户邻域模型与矩阵分解的推荐算法  

Recommendation Algorithm Based on User Nearest Neighbor Model and Matrix Factorization

作　　者：丁建军[1] 刘建平[1] 

机构地区：[1]浙江理工大学信息学院,浙江杭州310018

出　　处：《工业控制计算机》2015年第3期142-143,146,共3页Industrial Control Computer

摘　　要：针对协同过滤推荐算法所面对的稀疏矩阵和新用户问题,提出基于用户邻域模型与矩阵分解的推荐算法。通过对用户历史信息构建邻域模型以提高对新用户预测的准确性,同时考虑到矩阵稀疏和数据量较大会引起时间和空间复杂度过高,运用奇异值矩阵分解的方法,从而减小矩阵稀疏和数据量大的影响,提高推荐系统的准确性。通过Movie Lens数据集验证该算法的有效性。Concerning the difficulty of sparse matrix and new users problems in collaborative filtering recommendation algorithm,this paper proposes a new recommendation algorithm based on user nearest neighbor model and matrix factorization.To guarantee the prediction accuracy of new users,the nearest neighbor model based on user data and profile information is used.Meanwhile,large data sets and matrix sparsity would significantly increase time and space complexity.

关 键 词：电子商务 协同过滤 个性化推荐算法 矩阵分解 用户邻域模型 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]