In current teaching practices, teachers assess students’ mastery of specific knowledge points through quizzes and in-class questioning, which makes it difficult to provide targeted learning and hinders students’ personalized development. In response, this paper proposes an Ability Point Tracking Model (APTM) based on an exploration of Bayesian Knowledge Tracking (BKT) and Deep Knowledge Tracking (DKT). This model uses neural networks to encode students’ learning behaviors and predicts student performance through the dot product of ability indicator vectors and student state vectors. Compared to models like BKT and DKT, the APTM model offers greater interpretability. The APTM model was applied in practice using orchestral instrument teaching content. By analyzing students’ test responses on the day they completed the knowledge, one week later, and one month later, the model assesses students’ knowledge mastery and changes over time, generating diagnostic reports from both class and individual perspectives. The diagnostic structure effectively helps students identify their strengths and weaknesses and assists teachers in implementing differentiated instruction.
