Abstract: Collaborative filtering (CF)is the most successful recommended system to date, but traditional CF algorithm does not consider the problem of drifting users’ interests which often results in poor recommendation when users’ interests are changed. This paper develops a lineal gradual forgetting CF algorithm inspired by the forgetting law of psychology and it diminishes the importance of each rate with time. The experiment using MovieLens dataset shows that the new algorithm is more accurate than traditional CF algorithm.

Key words: Collaborative filtering, Interest drift, Lineal gradual forgetting

摘要： 协同过滤系统是目前最成功的一种推荐系统，但是传统的协同过滤算法没有考虑用户兴趣变化问题，导致用户兴趣发生变化时的推荐质量较差。该文借鉴心理学遗忘规律，提出了线性逐步遗忘协同过滤算法。该算法依据评价时间线性逐步减小每项评分的重要性。基于MovieLens数据集的实验结果表明，该算法在准确性方面优于传统的协同过滤算法。

关键词: 协同过滤, 兴趣变化, 线性逐步遗忘

CLC Number: 

• TP311