Article: A multimodal data fusion framework for online collaborative English learning interaction pattern analysis Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.21 pp.1 - 13 Abstract: This study proposes a multimodal artificial neural network - convolutional neural network (ANN-CNN) framework for the analysis of online collaborative English learning interaction patterns. This framework combines the original behavioural features with the frequency domain feature generated by the fast Fourier transform (FFT). The ANN branch focuses on learning direct behavioural relationships from original learning indicators, such as engagement, navigation, interaction, and assessment-related features. In contrast, the 1D-CNN branch learns hidden structural and fluctuation patterns from FFT-based frequency-domain features. Finally, the outputs of the two branches are fused for the final learning outcome prediction. Experimental results show that, compared with several traditional machine learning models including random forest, KNN, and decision tree, the proposed framework performs better. The accuracy of this multimodal ANN-CNN framework reaches 0.9600, and the F1 score reaches 0.9599. Moreover, other analysis, e.g., confusion matrix, further validates the strong discrimination ability of the proposed framework. Inderscience Publishers - linking academia, business and industry through research

Title: A multimodal data fusion framework for online collaborative English learning interaction pattern analysis

Authors: Baoqin Yan

Addresses: Fuzhou University Zhicheng College, No. 50 West Yangqiao Road, Hongshan Town, Gulou District, Fuzhou City, Fujian Province, 350002, China

Abstract: This study proposes a multimodal artificial neural network - convolutional neural network (ANN-CNN) framework for the analysis of online collaborative English learning interaction patterns. This framework combines the original behavioural features with the frequency domain feature generated by the fast Fourier transform (FFT). The ANN branch focuses on learning direct behavioural relationships from original learning indicators, such as engagement, navigation, interaction, and assessment-related features. In contrast, the 1D-CNN branch learns hidden structural and fluctuation patterns from FFT-based frequency-domain features. Finally, the outputs of the two branches are fused for the final learning outcome prediction. Experimental results show that, compared with several traditional machine learning models including random forest, KNN, and decision tree, the proposed framework performs better. The accuracy of this multimodal ANN-CNN framework reaches 0.9600, and the F1 score reaches 0.9599. Moreover, other analysis, e.g., confusion matrix, further validates the strong discrimination ability of the proposed framework.

Keywords: online collaborative English learning; multimodal learning; educational data mining; frequency-domain analysis; fast Fourier transform; FFT.

DOI: 10.1504/IJRIS.2026.155788

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.21, pp.1 - 13

Received: 20 May 2026
Accepted: 29 Jun 2026
Published online: 13 Aug 2026 *

Title: A multimodal data fusion framework for online collaborative English learning interaction pattern analysis

Keep up-to-date