原稿----一种基于机器学习的加权SlopeOne算法改进.docVIP

一种基于机器学习的加权Slope One算法改进 1，2，1，3，1，2 1 (燕山大学信息科学与工程学院,河北秦皇岛066004) 2 (燕山大学信息科学与工程学院计算机教学实验中心,河北秦皇岛066004) 3 (河北省计算机虚拟技术与系统集成重点实验室,河北秦皇岛066004) E-mail ：huansisi_ysu@163.com 摘　要：Slope One算法是一种基于内存的协同过滤推荐算法但在计算时，内存过大预测结果准确度不高，用户数量和项目数量，。本文将基于模型基于内存的Slope One算法，提出Slope One算法最后，在Filmtrust和Movielens数据集上的对比结果表明，融合偏差因子的加权Slope One算法在这两个稀疏度不同的数据集下，均能获得较高的推荐准确度。 关键词：协同过滤；；最小二乘法；Slope One算法 中图分类号：TP 　　　文献标识码：A　　　　　　　　　　文章编号：Improved Weighted Slope One Algorithm Based?on?Machine?Learning ZHANG Yu-lian1,2，HUAN Si-si1,3， LIANG Shun-pan1,2 1 (Information Science and Engineering College, Yanshan University, Qinhuangdao 066004, China) 2 (The Basic Computer Teaching Experiment Center, Information Science and Engineering College, Yanshan University, Qinhuangdao 066004, China) 3 (The Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province, Qinhuangdao 066004, China) Abstract: Slope One algorithm is a memory-based collaborative filtering recommendation algorithm, but in computing, the algorithm’s memory consumption is too large, and prediction accuracy is low, especially when the user and item number is larger. Therefore, this article puts forward integrate model-based and memory-based methods together ,called improved weighted Slope One algorithm based?on?machine?learning, the algorithm is intuitive and efficient, can overcome the traditional disadvantages of the memory-based recommendation algorithms. Finally, experiments on the well-known datasets Filmtrust and Movielens show that the biased integrating least squares technique into weighted Slope One algorithm in the case of different sparse datasets can both achieve great improvement of prediction accuracy. Key words: collaborative filtering;machine learning; least squares technique ;weighted Slope One algorithm1　引言 随着的发展和普及，信息过载问题愈加严重。在面对海量数据时，用户如何从冗余的信息中找到自己需要的内容，便成了一个值得研究的课题[1]。为有效解决信息过滤问题，推荐系统应运而生。协同过滤(Collaborative Filtering, CF)算法是推荐系统中较常用的算法之一[2]，是目前发展的最成熟的推荐技术，它仅仅通过分析用户的历史行为便可为