Knowledge tracing based on multi-feature fusion

Knowledge tracing involves modeling student knowledge states over time so that we can accurately predict student performance in future interactions and recommend personalized student learning paths. However, existing methods, such as deep knowledge tracing and dynamic key-value memory networks (DKVMN), fail to comprehensively consider some key features that may influence the prediction results of knowledge tracing. To solve this problem, we propose a new model called knowledge tracing based on multi-feature fusion (KTMFF), which introduces features of the question text, the knowledge point difficulty, the student ability, and the duration time, etc., provides feature extraction methods, and uses a multi-head self-attention mechanism to combine the above features. This model predicts student mastery levels of knowledge points more accurately. Experiments show that the area under curve (AUC) of the KTMFF model is 3.06% higher than that of the DKVMN model. Furthermore, the ablation study indicates that each of the above features can improve the AUC of the model.