Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (5): 137-141.

Previous Articles     Next Articles

Collaborative filtering algorithm based on improved nearest neighbors

SHUO Liangxun1, CHAI Bianfang2, ZHANG Xindong3   

  1. 1.Network Security Laboratory, Shijiazhuang University of Economics, Shijiazhuang 050031, China
    2.Institute of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044, China
    3.Hebei Xinji Network Media Co., LTD, Shijiazhuang 050031, China
  • Online:2015-03-01 Published:2015-04-08

基于改进最近邻的协同过滤推荐算法

硕良勋1,柴变芳2,张新东3   

  1. 1.石家庄经济学院 网络信息安全实验室,石家庄 050031
    2.北京交通大学 计算机与信息技术学院,北京 100044
    3.河北新冀网络传媒有限公司,石家庄 050031

Abstract: Aiming to the problems that the quality and precision are caused by the sparse user scorings and cold-start, a novel collaborative filtering algorithm based on improved nearest neighbors is proposed in this paper. User-item matrix is established, and similarity between items and users is measured, the nearest neighbor of items and users is acquired, in which the particle swarm optimization algorithm is used to select the optimal value of the parameter k, the simulation experiments are carried out on MovieLens and Book-Crossing dataset. The results show that the proposed algorithm can achieve lower MAE and efficiently improve recommendation precision, and it can enhance the quality of recommendations.

Key words: collaborative filtering, improved nearest neighbor, particle swarm optimization algorithm, selecting parameters

摘要: 针对当前协同过滤推荐算法易受数据稀疏性与冷启动的问题,提出了一种改进最近邻的协同过滤推荐算法。建立用户-项目评分矩阵,并度量项目之间、用户之间的相似性,获取项目和用户的最近邻居,其中最近邻居的最优参数k值采用粒子群算法选择,在MovieLens和Book-Crossing数据集上进行了仿真对比实验。结果表明,相对于其他协同过滤推荐算法,该算法降低了平均绝对误差值,提升了推荐准确度,达到提高推荐质量效果的目的。

关键词: 协同过滤, 改进最近邻, 粒子群优化算法, 参数选择