计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (3): 112-118.DOI: 10.3778/j.issn.1002-8331.2012-0467

• 大数据与云计算 • 上一篇    下一篇

融合时间加权信任与用户偏好的协同过滤算法

张岐山,朱猛   

  1. 福州大学 经济与管理学院,福州 350108
  • 出版日期:2022-02-01 发布日期:2022-01-28

Collaborative Filtering Algorithm Combining Time-Weighted Trust and User Preferences

ZHANG Qishan, ZHU Meng   

  1. School of Economics and Management, Fuzhou University, Fuzhou 350108, China
  • Online:2022-02-01 Published:2022-01-28

摘要: 针对现有的协同过滤推荐算法中存在评分数据稀疏和用户兴趣动态变化的问题,提出了融合时间加权信任与用户偏好的协同过滤算法。考虑到用户评分时间的不均匀,对时间权重进行改进,并将其融入到直接信任计算中,缓解用户兴趣动态变化的问题。通过信任传递得到的间接信任以及建立用户对项目标签的偏好矩阵得到用户之间的偏好相似度来缓解数据的稀疏性。融合用户的信任度与偏好相似度进行推荐。实验结果表明,与其他基准算法相比,提出的算法具有更高的F1值,提高了推荐质量。

关键词: 协同过滤, 时间权重, 信任, 项目标签, 用户偏好

Abstract: Aiming at the problems of sparse rating data and dynamic changes of user interests in the existing collaborative filtering recommendation algorithms, a collaborative filtering algorithm combining time-weighted trust and user preferences is proposed. Considering the unevenness of user rating time, the time weight is improved and incorporated into the direct trust calculation to alleviate the problem of dynamic changes in user interest. Through the indirect trust obtained by trust transfer and the establishment of user preference matrix for item tags, the preference similarity between users is obtained to alleviate the sparseness of data. The user’s trust and preference similarity are combined for recommendation. Experimental results show that compared with other baseline algorithms, the proposed algorithm has a higher F1 value, which improves the quality of recommendation.

Key words: collaborative filtering, time-weighted, trust, item tag, user preferences