Abstract: Set in technology innovation platform, traditional collaborative filtering algorithm which has the problem of lagged recommendation and bad scalability is focused on. According to users’ real-time feedback, incremental update mechanism is introduced, which detaches fixed factor and incremental factor. When the score is changed, only incremental factor needs to be changed, which improves the scalability of the algorithm and reflects the users’ interest change more accurately. The experimental results indicate that the algorithm ensures accuracy of the recommendation and at the same time reduces the recommendation time.

Key words: management science and engineering, incremental updating mechanism, slope one algorithm, user behavior, real-time recommendation

摘要： 以技术创新平台为背景,针对原有协同过滤算法推荐滞后以及算法可扩展性差的问题,根据用户的实时反馈,在Slope One算法的基础上,提出了更新增量机制,分解出固定因子以及增量因子,当用户对项目的评分改变时,只需更新增量因子,提高了算法的可扩展性,更精确地反应了用户的兴趣变化。经算例验证,该算法在保证推荐精度的同时可以有效地缩短推荐时间。

关键词: 管理科学与工程, 增量更新机制, Slope One 算法, 用户行为, 实时推荐