Predicting E-Learning Continuance Intention: Cross-Dataset Validation and Educational Intervention

The expansion of e-learning in higher education has increased the importance of understanding students' continuance intention (CI) after initial adoption. This study developed and externally validated a machine learning approach for predicting e-learning continuance intention and translated the predictions into a data-driven educational intervention framework. Dataset B (n=۳۶۸) was used for model development and Dataset EFL (n=۴۳۵) for independent external validation. Three common constructs-Expectation Confirmation (CON), Satisfaction (SAT), and CI- were harmonized across datasets. Five algorithms were compared using cross-validation, with CON and SAT as predictors and CI as the target. Linear Regression was selected based on the lowest mean cross-validation RMSE (۰.۴۷۲۱). On the internal test set, the model achieved MAE=۰.۳۲۵۴, RMSE=۰.۵۳۱۵, and R۲=۰.۵۹۴۱. Without retraining or retuning, external validation on Dataset EFL yielded MAE ۰.۲۲۳۵, RMSE=۰.۳۵۷۳, and R²=۰.۵۷۰۵. Sensitivity analyses showed that R² remained between ۰.۵۴۹۵ and ۰.۵۷۰۵ after quality-based filtering. Adding Perceived Usefulness, Perceived Educational Support, and Perceived Emotional Support did not produce a stable improvement across algorithm families. The final framework uses Predicted CI together with CON and SAT as decision-support signals for prioritization, human review, follow-up, and reassessment. The findings indicate that a simple, interpretable model can retain substantial predictive performance across independent e-learning datasets while supporting responsible, human-supervised educational decision-making.

کلیدواژه ها:

E-learning ، Continuance Intention ، Machine Learning ، External Validation ، Learning Analytics ، Educational Interventions