K-近邻矩阵分解推荐系统算法

K-nearest Neighbor Matrix Factorization for Recommender Systems

协同过滤算法随着电子商务的发展而提出,用来为电商用户作出智能推荐.近几年来,电子商务网站迅速发展,对协同过滤算法有更高的要求.传统的矩阵分解与K最近邻对用户评分矩阵整体进行学习从而达到推荐目的.事实上,用户评分矩阵有很高的稀疏性,有用的评分信息是与目标用户相关联的评分,计算用户对项目的评分,单一查找用户近邻与项目近邻并不符合实际应用.为了缓解稀疏性,使推荐结果更加合理,提出一个近邻矩阵分解算法,将用户近邻与项目近邻评分信息融合为一个近邻评分矩阵,挖掘目标用户对目标项目的评分信息.在真实数据集上的实验表明,提出的算法提高了推荐结果的准确性.

With the development of e-commerce site,the Collaborative Filtering was proposed to make intelligent recommendations for users.In recent years,the rapid development of e-commerce has a higher request for Collaborative Filtering algorithm.The traditional Matrix Factorization and K-Nearest Neighbor achieve recommendations by learning the user rating matrix.In fact,the user rating matrix is very sparse and useful rating information is associated with the target users.In addition,it is not practical to calculate rating only by finding the User＇s neighbor or Item＇s neighbor.This paper proposed a K-Nearest Neighbor Matrix Factorization to alleviate the sparse of rating matrix and to make recommendation more reasonable.The experimental results on real data sets show that the new algorithm can efficiently improve the recommendation accuracy.