In the context of personalized education, it remains challenging to understand each student’s mastery of knowledge points. Consequently, providing effective support for students’ weak areas of knowledge still needs to be addressed. To address this issue, this paper proposes a guided personalized learning assistance method. First, a NeuralCD-SEKA cognitive diagnostic model is constructed. By analysing the associations among students, exercises, knowledge points, and the time students spend on exercises, the model accurately assesses students’ knowledge mastery. Second, on the basis of students’ knowledge mastery and the importance of knowledge points, a personalized cognitive map is visualized to clarify learning pathways. Finally, an NAD-KR \(_{ADFM}\) knowledge point recommendation model is developed. This model precisely recommends suitable knowledge points to students on the basis of their mastery levels and the importance of the knowledge points. This model also uses a large language model to generate questions for guided questioning. The experimental results show that the proposed models achieve significant performance improvements in both cognitive diagnosis and knowledge point recommendation. It also shows that the recommended knowledge points are evenly distributed, providing reliable support for personalized learning.