关键词: 个性化网站, 基于用户的协同过滤算法, 推荐系统, 用户意图, 增量式更新

Abstract: In order to solve the problems that user’s retrieval intention is rarely considered in the personalized websites and the search result is not good, this paper proposes an Incremental Collaborative Filtering Recommendation method（ICFR）. The ICFR model improves the collaborative filtering recommendation algorithm, which is one of the most popular recommendation algorithms, and applies it to the personalized websites. Firstly, the browsing behavior information of users is extracted by analyzing web logs and normalized into rating value. Secondly, the similarity value between users is obtained by using the improved similarity calculation method, and the nearest neighbor set which can reflect the user’s intention is selected. Thirdly, the results predicted by the nearest neighbor set are sorted and returned to the user as the recommended list. Finally, historical user preference data are updated effectively in real time by the incremental updating algorithm. The experimental results indicate that the incremental collaborative filtering recommendation model is suitable for personalized websites and this method can make the retrieval results reflect the user intention.

Key words: personalized website, user-based collaborative filtering, recommendation system, user intention, incremental updating