Enhancing Machine Learning Algorithms Using Recursive Feature Elimination For Student Performance Prediction

Prediction systems for students' performance are crucial in improving educational outcomes in higher education institutions. Big data can be sourced from various channels, such as examination centers, academic systems, registration departments, and e-learning platforms. Extracting meaningful insights from educational data is challenging, necessitating data dimensionality reduction. This study introduces an enhanced recursive feature elimination (RFE) algorithm as a wrapper feature selection technique. We adopt machine learning by combining the proposed RFE with five different classifiers, including Gradient Boosting Classifier (GB), Random Forest Classifier (RF), Support Vector Classifier (SVC), K-Neighbors Classifier (KNN), and Decision Tree Classifier (DT). The dataset was originally intended to influence factors in dropout students, and it has 329 records and 9 features. The AUC score improved by 8% compared to the model tested without the proposed feature selection method by testing the proposed methodology with the Decision Tree Classifier. Educators and academic institutions can utilize this information to develop targeted intervention strategies, tailored learning experiences, and personalized recommendations for students, ultimately promoting a more effective learning environment and improving overall educational outcomes.