基于冲浪模型的Web协作推荐系统

WWW COLLABORATIVE RECOMMENDER SYSTEM BASED ON SURFING MODEL

当前基于协作过滤(CF,Collaborative Filtering)的推荐系统广泛应用于在线购物、音乐点播、智能Web推荐等系统。基于协作过滤的Web推荐系统的一个问题是用户通常仅仅访问很少Web页,因此根据用户访问Web页的记录找到一组相似用户的概率很低,这就是"稀疏问题"。本文提出了一种利用WWW冲浪模型,模仿用户访问Web页过程中的一些特点,并将用户的冲浪过程延续,模拟用户在Web站点访问更多的Web页,从而估计出用户对更多Web页的评价。本文还给出了实验比较,表明扩展冲浪深度后,系统推荐Web页的效果得到明显提高。

Collaborative Filtering has been used fairly successfully to build recommender systems in various domains, such as intelligent WWW navigators. However, these systems suffer from the sparsity: most user don't rate or visit most pages and hence the user-page matrix is typically very sparse, therefore the probability of finding a set of users with significant similar ratings or visitings is usually low. Our approach overcome this drawback, by utilizing the regularity of the user's previous visitings or ratings, virtually extending the user's surfing length in the Web site, so more document has been rated or visited, and thus we can get a compact user-page matrix. We present experimental resuets that show how this approach performs better than the former systems.