计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (3): 127-134.DOI: 10.3778/j.issn.1002-8331.2101-0029

• 大数据与云计算 • 上一篇    下一篇

基于子空间聚类的协同过滤推荐算法

王英博,韩国淼,王铭泽   

  1. 1.辽宁工程技术大学 创新实践学院,辽宁 阜新 123000
    2.辽宁工程技术大学 软件学院,辽宁 葫芦岛 125105
    3.南宁学院 会计学院,南宁 530200
  • 出版日期:2022-02-01 发布日期:2022-01-28

Collaborative Filtering Recommendation Algorithm Based on Subspace Clustering

WANG Yingbo, HAN Guomiao, WANG Mingze   

  1. 1.School of Innovation and Practice, Liaoning Technical University, Fuxin, Liaoning 123000, China
    2.School of Software, Liaoning Technical University, Huludao, Liaoning 125105, China
    3.Accounting College of Nanning University, Nanning 530200, China
  • Online:2022-02-01 Published:2022-01-28

摘要: 为了降低数据稀疏性对推荐算法效率产生的影响,提出一种基于子空间聚类的协同过滤推荐算法(SCUCF)。该算法创建感兴趣、不感兴趣以及既不感兴趣也不不感兴趣三种类型被评价项目的不同子空间。利用项目子空间为目标用户绘制邻居用户树,以此来寻找目标用户的邻居。利用改进的用户相似性计算方法来确定推荐用户。通过MovieLens 100K、MovieLens 1M数据集对算法进行了验证,实验结果表明,该算法能够使推荐算法的推荐性能得到提升。并且,在与其他同类改进算法相比,该算法也表现出一定的优越性。

关键词: 协同过滤, 子空间聚类, 邻居用户树, 相似性

Abstract: In order to reduce the impact of data sparsity on the efficiency of the recommendation algorithm, a collaborative filtering recommendation algorithm based on subspace clustering(SCUCF) is proposed. The algorithm creates different subspaces of three types of evaluated items of interest, not interest, and neither interest nor interest. Using the project subspace to draw a neighbor user tree for the target user to find the neighbors of the target user. An improved user similarity calculation method is used to determine recommended users. The algorithm is verified by MovieLens 100K and MovieLens 1M data sets. Experimental results show that the algorithm can improve the recommendation performance of the recommendation algorithm. Moreover, compared with other similar improved algorithms, this algorithm also shows certain advantages.

Key words: collaborative filtering, subspace clustering, neighbor user tree, similarity