Student success prediction using classification methods based on ensemble techniques and genetic algorithm feature selection

Educational data mining has gained popularity in recent years, providing strategies for extracting knowledge that might improve teaching methods and student outcomes. Feature selection, which determines the most pertinent features for the task, is a crucial component of creating precise and effective prediction models in this area. To do this, we evaluated various classification methods for predicting academic success and used the Genetic Algorithm (GA) to choose the best features for an educational dataset from the Kalboard360 Learning Management System (LMS). Behavioral features are successfully identified by the GA method as important features of student performance. Using ensemble techniques like boosting, bagging, and stacking, we further enhanced the accuracy results from the classifiers we tested, which included Decision Trees (DT), Naïve Bayes (NB), Logistic Regression (LR), Artificial Neural Networks (ANN), and K-Nearest Neighbor (KNN).The results show that a Gaussian Naïve Bayes (GNB) model, combined with a bagging ensemble, achieving the highest accuracy of 87.5%, demonstrating its reliability for predicting academic outcomes.