摘要
针对传统协同过滤推荐算法(CF)存在用户-项目矩阵稀疏以及推荐准确率较低等问题,提出了一种基于NKL和K-means聚类的协同过滤推荐算法(NKL-KM).首先,NKL-KM算法定义了一种新的相似性度量方法,该方法在进行相似性度量时考虑了各项目评分的分布以及评分值差异.其次,NKL-KM算法将K-means算法与CF算法结合,提高了推荐算法精度.最后,在MovieLens和Netflix数据集上进行算法对比实验,实验结果表明NKL-KM算法有较高的推荐精度.
A collaborative filtering recommendation algorithm(NKL-KM)based on NKL and K-means clustering is proposed to solve the problem of sparse user-item matrix and low recommendation accuracy in the traditional collaborative filtering recommendation algorithm(CF).Firstly,a new similarity measure method is defined,which takes into account the distribution of each item’s scores and the difference of score values.Secondly,the K-means algorithm is combined with the CF algorithm to improve the accuracy of recommendation algorithm.Finally,the algorithm comparison experiments are performed on MovieLens and Netfix datasets.It is proved that NKL-KM algorithm has higher recommendation accuracy.
作者
李顺勇
张钰嘉
张海玉
LI Shunyong;ZHANG Yujia;ZHANG Haiyu(School of Mathematical Sciences,Shanxi University,Taiyuan 030006,China;School of Journalism,Shanxi University,Taiyuan 030006,China)
出处
《河南科学》
2020年第1期6-12,共7页
Henan Science
基金
国家自然科学基金项目(61573229)
山西省基础研究计划项目(201701D121004)
山西省回国留学人员科研资助项目(2017-020)
山西省研究生教育改革项目(2019JG023)
太原市科技计划研发项目(2018140105000084)。
关键词
协同过滤
推荐算法
矩阵稀疏
K-MEANS
相似性度量
collaborative filtering
recommendation algorithm
sparse matrix
K-means
similarity measure
