JCSE, vol. 20, no. 2, pp.63-69, 2026

DOI: http://dx.doi.org/10.5626/JCSE.2026.20.2.63

Predicting Students’ Learning Activity on Online Education Platforms through Deep Learning

Nana Wang
School of Liberal Education, Liaoning University of International Business and Economics, Dalian, China

Abstract: On online education platforms, students’ learning activity is generally poor. Through certain predictive models, students’ learning activity can be predicted and intervened in. This paper implemented data balance processing with the k-means?synthetic minority oversampling technique (SMOTE) on the KDD Cup2015 dataset and then designed a bidirectional gated recurrent unit-attention (BiGRU-ATT) prediction model. Experiments found that the BiGRU-Att model showed the best performance in learning activity prediction compared with algorithms such as logistic regression and convolutional neural network, with an accuracy of 0.9126, an F1-score of 0.9420, and an area under the curve (AUC) value of 0.8997. After being treated with k-means-SMOTE, the accuracy of the BiGRU-Att model increased by 4.36%, its F1-score increased by 4.59%, and its AUC value increased by 9.39%. When data from different weeks were used as input for the BiGRU-Att model, the greater the number of weeks, the better the prediction effect. The results demonstrate the performance of the designed BiGRU-Att model in predicting learning activity, indicating that it can be applied to actual online education platforms.

Keyword: Deep learning; Online education platform; Prediction model; Learning activity

Full Paper:   7 Downloads, 6 View