A Knowledge Tracking Model Based on Students' Learning Abilities
Zhao Liu, Bin Liu, Zhonghua Cheng · 2024
To address the current lack of modeling students’ individual learning ability characteristics in knowledge tracing models, we propose a knowledge tracing model based on students’ learning ability. We hypothesize that students who master a knowledge point through fewer question practices demonstrate better learning ability. Based on this assumption, we design a Learning Ability-enhanced Sequential Neural Network module (LASNN), which incorporates learning ability as a crucial factor in the acquisition and updating process of students’ knowledge states. We employ an encoder-decoder structure based on self-attention mechanism to extract knowledge states relevant to the question to be predicted, and use GRU to capture sequential information in students’ question sequences. By combining these two types of information - students’ knowledge states and learning abilities - we predict students’ subsequent performance on questions. Our proposed model performs well on three real-world public datasets. Ablation experiments verify the effectiveness of each module, providing valuable insights for knowledge tracing models.