基于奇异值分解的slopeone算法.pdf

12  2012年第2卷 第11期 基于奇异值分解的Slope One 算法# 1,2 1,2** 林德军 ，孟祥武 （1. 北京邮电大学智能通信软件与多媒体北京市重点实验室，北京 100876； 2. 北京邮电大学计算机学院，北京 100876 ） 摘要：Slope One 算法是一种基于项目的协同过滤推荐算法，利用线性回归模型来预测评 分。Slope One 算法在数据稀疏的情况下推荐效果不好，而矩阵分解在抗稀疏性方面有很有 效，因此本文引入奇异值分解方法改进Slope One 算法。实验结果表明，该方法有效地解决 了Slope One 在数据的稀疏情况下推荐效果较差的问题，提高了推荐质量。 关键词：计算机应用；推荐系统；奇异值分解；Slope One 中图分类号：TP393 Slope One Algorithm Based on Single Value Decomposition Lin Dejun1,2, Meng Xiangwu1,2 (1. Beijing Key Laboratory of Intelligent Telecommunications Software and Multimedia, Beijing University of Posts and Telecommunications, Beijing, 100876, China; 2. Beijing University of Posts and Telecommunications, School of Computer Science, Beijing 100876) Abstract: Slope One algorithm is item-based collaborative filtering algorithm. It uses linear regression model to predict value. Slope One algorithm recommendation result is worse than the other CF algorithm under the condition of sparse data. However, matrix decomposition have great effect in sparseness resistance. In this paper, single value decomposition is used to improve Slope One algorithm. The experimental results show that Slope One algorithm based on single value decomposition sloves the problem that the effect of Slope One recommendation is worse, and improve the quality of recommendation. Key words: computer application; recommend system; Single Value Decomposition; Slope One 0 引言 随着互联网信息量的膨胀和电子商务的迅速发展，信息过载问题困扰着每个用户。推荐 系统作为信息过载的有效解决手段，能够缓