In view of the problem that the collaborative filtering algorithm ignores the learners' knowledge domain (learning state), this study improves the collaborative filtering algorithm used in the recommendation of personalized education. The recommendation algorithm is divided into three steps. (1) Based on cognitive diagnosis model, the study builds up a model construction analysis of the learner's knowledge domain based on learner's response matrix. (2) It Uses the collaborative filtering algorithm, combined with the knowledge domain of the target learners to analyze the learners with similar behaviors. (3) According to the similar learner's historical behaviors and the target learner's knowledge domain, the system would recommend testing questions (items) for the target learners. This recommendation method not only draws lessons from the generality of the similar learners of the same group, but also takes into account the uniqueness of the individual learners. The study combines the two to recommend the individualized items for the target learners, which ensures the accuracy and performance of the recommendation method. In the individualized education system, the recommendation method combined with cognitive diagnosis and collaborative filtering algorithm is an improved application.