A knowledge tracing based intelligent framework for formative assessment of students skills

Formative assessment, a crucial technique in personalized education, enables the dynamic tracking of a student’s knowledge state over time. By leveraging this method, educational systems can tailor learning experiences to individual needs, optimizing content delivery and assessments to enhance learning outcomes. In this research paper, we present an enhanced Deep Knowledge Tracing (DKT) model by incorporating Transformer-based architectures to improve the performance of knowledge tracing for formative assessment in educational settings. Traditional DKT models, which typically rely on Long Short-Term Memory (LSTM) networks, have demonstrated capabilities in tracking student’s knowledge states over time. However, these models often struggle with capturing long-term dependencies and complex patterns in students’ learning trajectories. To address these limitations, we propose using Transformer-based Deep Knowledge Tracing models, which offer superior performance in handling sequential data due to their self-attention mechanisms. The experimental results on three benchmark datasets reveal that the Transformer-based DKT model outperforms the LSTM-based counterpart in terms of AUC, accuracy, F1 score, and mean loss. The improved model shows a marked enhancement in identifying students’ knowledge gaps and predicting their future performance. This advancement facilitates more personalized and adaptive learning experiences and provides deeper insights into students’ learning processes. The findings highlight the potential of Transformer architectures in revolutionizing formative assessment methodologies, paving the way for more effective educational technologies.