Aiming at the problem of poor accuracy in tracking the knowledge status of students’ civic education, this study proposes a deep knowledge tracking model incorporating domain features with the support of educational knowledge graph, the DKT-KG model. The model filters important assessment behavioral features through a decision tree and incorporates the knowledge dependencies characterized by the educational knowledge graph to solve the problem of poor prediction accuracy of the original deep knowledge tracking model due to the lack of domain features. The experimental results on the ASSISTments09, Junyi, and KDD datasets show that the DKTKG model can more accurately track knowledge points of the mastery level, and the AUC index and F1 scores are higher than those of other comparison models. Empirical analysis was conducted using historical data of question answering in four Civics courses taken by learners on the online education platform. The average knowledge mastery probability and its corresponding prediction accuracy derived when learners do different amounts of questions show that the deep knowledge tracking model constructed in this paper can accurately predict students’ knowledge mastery.
