Abstract
Knowledge tracing models predict students' mastery of specific knowledge points by analyzing their historical learning performance. However, existing methods struggle with handling a large number of skills, data sparsity, learning differences, and complex skill correlations. To address these issues, we propose a knowledge tracing method based on dual graph convolutional networks and positive/negative feature enhancement. We construct dual graph structures with students and skills as nodes, respectively. The dual graph convolutional networks independently process the student and skill graphs, effectively resolving data sparsity and skill correlation challenges. By integrating positive/negative feature enhancement and spectral embedding clustering optimization modules, the model efficiently combines student and skill features, overcoming variations in learning performance. Experimental results on public datasets demonstrate that our proposed method outperforms existing approaches, showcasing significant advantages in handling complex learning data. This method provides new directions for educational data mining and personalized learning through innovative graph learning models and feature enhancement techniques.