| 摘 要: 协同过滤算法是实现推荐系统最重要的技术之一。随着时间的推移，用户对物品的偏好会不断地发生变 化，物品自身的流行度也会随时间不断地发生变化。目前常用的推荐算法如基于邻域的协同过滤算法itemCF、userCF 和隐语义模型算法FunkSVD、BiasSVD、SVD++都没有考虑到时间因素对推荐系统推荐质量的影响。而时间信息是一 种非常重要的上下文信息，应该在算法中加以利用。本文使用Sigmoid函数和流行度函数将时间因素融入到了BiasSVD 算法中，成功的设计出了一个融合时间信息的新算法Time-BiasSVD。在MovieLens数据集上的验证结果表明：该算法 与已有协同过滤算法，以及融合时间信息的算法timeSVD++相比，能更准确地预测用户实际评分，提高推荐系统的推 荐质量。 |
| 关键词: 偏置项；协同过滤；时间因素；Sigmoid函数；流行度函数 |
|
中图分类号: TP399
文献标识码: A
|
| 基金项目: 国家自然科学基金“在线评论偏差对商家信誉和平台收益的影响及纠偏措施研究”，批准号：71862027. |
|
| Collaborative Filtering Algorithm Based on Integrating Time Factors into Multiple Biases |
|
LI Mingxiu,WANG Shujun,JIA Ru,CHEN Lirong1,2
|
1.( 1.College of Computer Science & Technology, Inner Mongolia University, Huhhot 010000, Chna;2.
2.College of Computer Science & Technology, Tianjin University, Tianjin 300350, China)
|
| Abstract: Collaborative filtering algorithm is one of the most important techniques to implement recommendation system.As time goes by,users' preferences for items will change constantly and the popularity of items will also change over time.At present,the commonly used recommendation algorithms,such as neighborhood-based collaborative filtering algorithm itemCF,userCF and implicit semantic model algorithm FunkSVD,BiasSVD,SVD ++,do not take into account the impact of time factors on the recommendation quality of the recommendation system.In this paper,the Sigmoid function and the popularity function are used to design a new algorithm by integrating the time factor into the BiasSVD algorithm.The verification results on the MovieLens dataset show that the proposed algorithm can accurately predict the user's actual score and improve the recommendation quality of the recommendation system compared with the existing collaborative filtering algorithms and the timeSVD ++ algorithm. |
| Keywords: bias;time factor;collaborative filtering;Sigmoid function;popularity function |
