Course Learning Early Warning Model Based on Optimized Bagging-Stacking and Random Forest

The rapid development of educational digitization is driving the application of artificial intelligence technology in the field of education. In the course learning process, using machine learning algorithms based on formative assessment data to predict, on a rolling basis, whether students will pass the final course assessment and provide timely warnings is an effective approach to improving course learning outcomes. To enhance predictive performance, a Stacking ensemble learning model is proposed, utilizing tabu search optimized Bagging-Naive Bayes and random forest models as base learners. This model leverages continuously updated formative assessment data throughout the course learning process to achieve ongoing predictions and rolling warnings for course learning. Comparative results with other algorithms demonstrate that the proposed model significantly enhances the accuracy of learning warnings, with the highest improvement in recall on the test set being 20.5% and the lowest improvement being 2.6%. The accuracy of the model also far exceeds that of using a single algorithm.