摘要： 针对个性化服务的系统中，如何将新发现的用户兴趣和原有兴趣合并为用户的新兴趣的问题，提出了一种基于概念相关性的用户兴趣漂移方法。采用混合模型，将用户兴趣分为长期兴趣和短期兴趣，对短期兴趣采用滑动窗口处理更新，对长期兴趣采用基于概念相关的渐进遗忘方法，实验表明，该方法不仅能够较为准确地跟踪用户的兴趣变化，而且能够预测用户的兴趣，具有较好的效率。

关键词: 概念漂移；概念相关性；渐进遗忘；滑动窗口

Abstract: This paper aims at the problem of combination of user’s interests and previous interests. It proposes a hybrid model based on concept correlations. In this hybrid model, user’s interests consist of long-term model and short-term model, and these two models adopt different algorithms to deal with the interest’s changes. Short-term model uses sliding window algorithm to deal with the sequence of the user’s interest samples. Long-term model uses gradual forgetting algorithm based on concepts correlations to calculate the interests’ weight. Experiment shows that this method is more efficient; it can not only trace the interest change of the user exactly, but also predict the interest of the users.

Key words: Concept drift; Concept correlations; Gradual forgetting; Sliding window