Abstract: This paper describes a new model for collaborative filtering, it can be used to slove the problem that two users with similar preferences on items may have different rating schemes. This model may effectively improve the traditional collaborative filtering method used to compute the similarity between users, and enhances the accuracy of user similarity measurement. Experiment results show that the new model performs well in personalized recommendation system.

Key words: Collaborative filtering, Similiarity, Rating model, Preference model

摘要： 提出了一种新的协同过滤模型，解决了不同用户在项目上，有相似的偏好、不同的评分习惯的问题。该模型可有效地改进传统协同过滤模型相似性度量方法，提高了用户相似性度量准确性。实验结果表明，该模型在个性化推荐系统应用中取得了较好的效果。

关键词: 协同过滤, 相似性, 评价模型, 偏好模型