Framelet-based dual hypergraph neural networks for student performance prediction

In the field of educational data mining, accurately predicting student performance is vital for developing effective educational strategies. However, existing methods often fall short in capturing the complex relationships between students, focusing mainly on individual attributes. This paper introduces a pioneering framelet-based dual hypergraph neural network (FD-HGNN) model to advance this task. Our innovative approach decomposes student feature matrices into low-pass and high-pass components using a framelet-based transform. These components form the basis for creating hypergraphs, capturing intricate student relationships. The model integrates a dual hypergraph neural network with distinct channels for low-pass and high-pass components, augmented by a variance interaction layer employing an attention mechanism. This structure ensures a more comprehensive representation of student data, enhancing prediction accuracy. Extensive validation against traditional machine learning methods and graph neural networks across four real-world educational datasets demonstrates the superiority of our approach. The findings highlight the significant potential of our model in revolutionizing student performance prediction in educational settings.