Difficulty-aware Convolutional Knowledge Tracing for Student Performance Prediction
Yuan Zeng, Tiancheng Jin, Liang Dou · 2022
Predicting students’ performance in future learning activities is an important task in online education systems. It enables education researchers to understand students’ learning profiles and provide them with personalized teaching services based on this knowledge. Most existing knowledge tracing methods do not adequately model the difficulty of the exercise and instead model exercise information, student ability, and knowledge state features while ignoring their characteristics. To address these challenges, we propose a difficulty-aware CNN-based knowledge tracing model that extracts several factors effective in modeling exercise difficulty and models changes in students’ knowledge states and abilities at a microscopic level. Our model can simulate students’ learning processes more realistically and performs well on real-world datasets.