Integrating Forgetting Factor and Attention Mechanism Into Knowledge Tracing

Knowledge tracing (KT) is a core technology in smart education, providing critical information for applications such as exercise recommendation and personalized learning. Deep learning-based KT models, such as DKT, fail to consider forgetting factors, while the SAINT model accounts for forgetting factors but only as input features. This paper proposes a Forgetting Attention Knowledge Tracing (FAKT) model by quantifying multiple forgetting factors into a single forgetting factor and integrating it with attention mechanisms. FAKT demonstrates improved prediction accuracy across four public datasets.