A Deep Knowledge Tracking Model Incorporating the Forgotten Features of Multiple Concepts
Biao Kong, Dongfeng Liu · 2024
Knowledge tracking is based on the historical answer records of students to track the changes in their knowledge status. With the rapid development of deep neural networks, they are also widely used in the field of knowledge tracking. Most of the existing deep knowledge tracking models only consider the situation where the question contains only one concept, and cannot solve the problem where the question contains multiple concepts, and they do not consider the effect of forgetting behaviour on the knowledge state change during the learning process. A new deep knowledge tracking model is proposed to solve these problems. This model uses multiple concepts to express the problem and considers the time interval for students to learn the same concepts and the number of times they learn the same concepts as two important forgetting features of the deep knowledge tracking model. These forgetting features exert an influence in the prediction stage of the model and in the knowledge state change stage to increase the interpretability of the model. Experiments on publicly available datasets show that the model prediction process is more realistic and has better prediction accuracy.