基于改进K-means和优化评分的协同过滤推荐算法  

Collaborative filtering recommendation algorithm based on improved K-means and optimized scoring

作　　者：施天虎 徐洪珍[1] SHI Tianhu;XU Hongzhen(School of Information Engineering, East China University of Technology, Nanchang 330013,China)

机构地区：[1]东华理工大学信息工程学院,南昌330013

出　　处：《江苏科技大学学报（自然科学版）》2021年第6期72-77,共6页Journal of Jiangsu University of Science and Technology:Natural Science Edition

基　　金：国家自然科学基金资助项目(62066003);江西省教育厅科技计划项目(GJJ160554);江西省放射性地学大数据技术工程实验室开放项目(JELRGBDT201802)。

摘　　要：针对传统协同过滤推荐算法评分数据稀疏、没有考虑推荐的时效性而导致推荐准确性不佳的问题,提出了一种基于改进K-means和优化评分的用户协同过滤推荐算法,在评分矩阵中加入用户对项目类别的评分,并使用Weigh Slope One算法得到的预测评分替代评分矩阵中的未评分项,以此降低数据稀疏性;并改进K-means聚类算法,对填充后的用户数据进行聚类,引入时间权重,将时间因子纳入评分预测中.在MovieLens-100K数据集上进行仿真实验,实验结果表明:所提算法较好解决了评分数据稀疏,推荐时效性差的问题,且推荐效果具有明显提升.Aiming at the problem of poor recommendation accuracy due to sparse scoring data and regardless of the real-time recommendation in traditional collaborative filtering recommendation algorithms,a user collaborative filtering recommendation algorithm based on improved K-means and optimized scoring is proposed.Firstly,the user′s score of item category is added to the scoring matrix,and the predicted scores obtained by the Weigh Slope One algorithm are used to replace the unscored items in the scoring matrix to reduce the data sparsity;secondly the K-means clustering algorithm is improved to cluster the filled user data,the time weight is introduced,and the time factors are included in the score prediction.A simulation experiment was carried out on the MovieLens-100K data set.The experimental results show that the proposed algorithm can well solve the problems of sparse scoring data and poor real-time recommendation,and the recommendation effect has been significantly improved.

关 键 词：协同过滤 聚类 时间权重 优化评分 weight slope one算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]