This paper introduces the student-centered new intelligent, comprehensive teaching model. This paper uses the accumulated user behavior data on the platform to build a personalized recommendation model. The system provides a personalized learning route for students to choose learning strategies and resources. This method converts the user’s learning process into the user’s evaluation of the system to solve the problem of the scoring matrix. Secondly, it proposes to improve users’ similarity based on their initial marking to overcome the personalized model concept of cold start questioning of new customers effectively. Examples show that the proposed method can improve the efficiency of personalized recommendations.