基于NystrOm方法的电影推荐算法  

Movie recommendation algorithm based on NystrOm method

作　　者：杨美姣 刘惊雷 YANG Meijiao;LIU Jinglei(College of Computer and Control Engineering,Yantai University,Yantai 264005,China)

机构地区：[1]烟台大学计算机与控制工程学院,山东烟台264005

出　　处：《应用科技》2018年第4期82-88,共7页Applied Science and Technology

基　　金：国家自然科学基金项目(61572419,61773331,61703360);山东省高等学校科技计划项目(J17KA091)

摘　　要：针对传统推荐系统中推荐效率较低的问题,提出了一种与Nystr?m方法相结合的推荐系统。设计了一Nystr?m方法和非负矩阵分解(non-negative matrix factorization,NMF)相结合的推荐方法。即先用Nystr?m方法提取用户或电影的特征,然后用NMF对用户或电影的特征进行分析。提出的Nystr?m方法提取特征的算法解决了因矩阵规模较大发生溢出的问题,NMF方法能保证提取特征的精度,将2种方法相结合,不仅能够加快计算的速度,同时也能提高系统的推荐效率。最后通过真实的900个用户对1 500部电影的评分矩阵进行了测试,与其他算法相比,精度有了明显的改进。To solve the problem of low recommendation efficiency in traditional recommendation systems, this paper proposes a recommendation system combined with Nystrom method. A new recommendation algorithm was designed by combination of Nystrom method and non-negative matrix factorization（NMF）. That is, extract the characteristics of the user or movie first by Nystrom, and then analyze the characteristics of the user or the movie by NMF. The proposed Nystrom method for extracting features solves the problem of overflow due to large size of the matrix, while the NMF algorithm guarantees accuracy of the extracted features. Combining the two methods can not only speed up the calculation, but also improve recommendation efficiency of the system. Finally, 900 real users were tested on the scoring matrix of 1500 movies. Compared with the other algorithms, the accuracy was improved significantly.

关 键 词：推荐系统 NYSTROM方法 NMF 特征提取 精度 效率 矩阵溢出 评分矩阵 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]