Prediction of Students’ Performance in Flipped Classroom-based Vocational Education Using Data Mining Tools

Abstract

This study proposes a two-stage classification approach to predict students’ performance in a flipped classroom approach, using data from Moodle LMS, organized according to the Revised Bloom’s Taxonomy and the Cognitive Theory of Multimedia Learning. The approach involves classifying students as Pass or Fail in the first stage, and then further classifying Pass students into A, B, or C grades in the second stage. The findings indicate that applying the two-stage approach with the Synthetic Minority Over-sampling Technique (SMOTE) outperforms the one-stage approach for addressing the issue of multi-class imbalanced learning. Various supervised machine learning techniques were evaluated, including Support Vector Machines (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), and Gradient Boosting Trees (GBT), with GBT achieving the highest performance overall. KNN also showed the most significant improvement from the traditional one-stage to the two-stage classification. This method enhances the accuracy of identifying at-risk students and predicting their final grades, which enhances early warning systems and personalized learning interventions in flipped classrooms.