Abstract: To introduce modified collaborative filtering recommendation algorithm to knowledge recommendation system, the user cluster was carried out by calculating the similarity of learning objectives, background and cognitive competence between the different learners and then completing recommendation using collaborative filtering algorithm in the user cluster. The results showed that this method could improve the accuracy and efficiency of recommendation, and could meet user's meeds. In the process of knowledge recommendation, it also exits new user problem, real-time problem similar to traditional collaborative filtering recommendation. The specific solution method will be the main content of future research.