计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (11): 161-165.DOI: 10.3778/j.issn.1002-8331.1701-0332

• 模式识别与人工智能 • 上一篇    下一篇

引入兴趣稳定性的时间敏感协同过滤算法

张  旭,孙福振,方  春,郭  蕊   

  1. 山东理工大学 计算机科学与技术学院,山东 淄博 255049
  • 出版日期:2018-06-01 发布日期:2018-06-14

Collaborative filtering algorithm with stability of interest and time-sensitive

ZHANG Xu, SUN Fuzhen, FANG Chun, GUO Rui   

  1. College of Computer Science and Technology, Shandong University of Technology, Zibo, Shandong 255049, China
  • Online:2018-06-01 Published:2018-06-14

摘要: 针对物品流行偏置现象,将物品流行度引入到用户兴趣中建模,提出了基于物品流行度的用户兴趣特征相似度模型。针对传统模型没有考虑到用户兴趣稳定性和难以实时捕获用户兴趣问题,在计算用户兴趣相似度过程中引入时间敏感和兴趣稳定性,提出了引入时间敏感的用户兴趣稳定性的相似度模型。最后,融合权重因子将两种相似度模型线性加权,提出了一种新颖的IPSTS算法模型。实验表明,该算法相比传统协同过滤算法在平均绝对误差(MAE)和均方根误差(RMSE)上均有明显降低的同时也能挖掘出长尾物品,缓解了物品流行偏置现象。

关键词: 时间敏感, 兴趣稳定性, 物品流行度, 协同过滤, 流行偏置

Abstract: In terms of the phenomenon of object popularity bias, a user interest feature similarity model based on item popularity is proposed by introducing object popularity into user interest modeling. What’s more, in terms of  the problem that the traditional model does not take the stability of user interest and the difficulty of capturing user interest in real time into consideration, a similarity model with time-sensitive and interest-stability is proposed, which is based on introducing time-sensitive and interest-stability during calculating user interest similarity. Finally, a novel IPSTS algorithm model is proposed by using the weighting factors to linearly weight the two similarity models. Experiments show that the algorithm can significantly reduce the Mean Absolute Error(MAE) and Root Mean Square Error(RMSE) compared with the traditional collaborative filtering algorithm, and can also excavate long tail items and relieve phenomenon of popularity bias.

Key words: time-sensitive, stability of interest, prevalence of item, collaborative filtering, popularity bias