Multimodal graph learning based on 3D Haar semi-tight framelet for student engagement prediction

Ming Li, Xiaosheng Zhuang, Lu Bai, Weiping Ding

Abstract: Highlights•A novel multimodal graph learning framework for student engagement prediction is proposed.•3D Haar semi-tight framelet transform is introduced to facilitate multimodal data fusion.•Spectrum-based graph structure learning is proposed to build multimodal graph.•Our framework outperforms several baselines on a real-world multimodal educational dataset.