Article: Multimodal learning behaviour clustering and psychological cognitive state assessment algorithm Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.13 pp.40 - 57 Abstract: Accurate assessment of psychological and cognitive states is a core requirement for personalised education. However, existing research has not sufficiently considered multimodal learning behaviour characteristics, resulting in low evaluation accuracy. To address this, this paper first introduces hypergraph contrastive autoencoders to capture higher-order correlations among multimodal learning behaviours, incorporating structural information into end-to-end clustering to enhance its robustness. Building upon this, the clustering results of learning behaviours are used as input to further design a psychological cognitive state assessment algorithm based on multimodal learning behaviour clustering. Hypergraph neural networks are employed to model complex interactions among multimodal learning behaviour features, with decoupling methods used to separately extract intrinsic and extrinsic feature representations. Experiments on the KDD Cup 2010 education dataset demonstrate that the proposed algorithm achieves clustering purity and evaluation F1 scores of 94.38% and 94.95%, respectively, significantly outperforming comparison methods. Inderscience Publishers - linking academia, business and industry through research

Title: Multimodal learning behaviour clustering and psychological cognitive state assessment algorithm

Authors: Conghui Liu

Addresses: Department of Preschool Education, Zhengzhou Institute of Technology, Henan, 450000, China

Abstract: Accurate assessment of psychological and cognitive states is a core requirement for personalised education. However, existing research has not sufficiently considered multimodal learning behaviour characteristics, resulting in low evaluation accuracy. To address this, this paper first introduces hypergraph contrastive autoencoders to capture higher-order correlations among multimodal learning behaviours, incorporating structural information into end-to-end clustering to enhance its robustness. Building upon this, the clustering results of learning behaviours are used as input to further design a psychological cognitive state assessment algorithm based on multimodal learning behaviour clustering. Hypergraph neural networks are employed to model complex interactions among multimodal learning behaviour features, with decoupling methods used to separately extract intrinsic and extrinsic feature representations. Experiments on the KDD Cup 2010 education dataset demonstrate that the proposed algorithm achieves clustering purity and evaluation F1 scores of 94.38% and 94.95%, respectively, significantly outperforming comparison methods.

Keywords: learning behaviour clustering; autoencoder; feature decoupling; hypergraph neural network.

DOI: 10.1504/IJRIS.2026.154057

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.13, pp.40 - 57

Received: 15 Jan 2026
Accepted: 15 Feb 2026
Published online: 10 Jun 2026 *

Title: Multimodal learning behaviour clustering and psychological cognitive state assessment algorithm

Keep up-to-date