中文摘要：

在用户的协同过滤推荐模型中，用户对项目评分的偏好行为会导致计算用户之间的相似性出现偏差，影响推荐的质量。文章根据用户的评分习惯划分用户，利用大间隔寻找用户的近似邻居，提出了一种基于用户偏好的协同过滤推荐算法，首先引入一种新的相似性度量方法计算用户之间的相似度，再构建一种基于用户偏好的协同过滤推荐模型。实验结果表明，该算法能有效提高推荐质量。

英文摘要：

Users have different rating preference in the model of the user-based collaborative filtering recommendation, and the preference behavior leads to the deviation of calculating the similarity among users. Consequently, the recommendation quality of systems is restricted, On this basis, all users are divided into different groups according to user＇s preference behavior, and the method of large margin is presented to define user＇s neighborhood, and an algorithm of collaborative filtering recommendation based on user preference is proposed. Firstly, the similarity among users is calculated by introducing a new similarity measure method. Then a model of collaborative filtering recommendation based on user preference is constructed. Finally, the experimental results show that the proposed algorithm can im- prove the recommendation quality effectively.