A Comparative Analysis of Ensemble and Optimized Machine Learning Models for Multiclass Student Performance Prediction

Predicting student educational performance is an indispensable research area in educational data mining and learning analytics, contributing data-driven insights for early interventions and academic planning. This study applied multiple machine learning algorithms to classify student performance into four levels Poor, Average, Good, and Excellent using the enhanced_student_habits_performance_dataset.csv from Kaggle, which contains 80,000 records describing academic scores, study habits, family background, socio-economic status, and health factors. Eight models were implemented and evaluated in Python using scikit-learn, including Random Forest, Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naive Bayes, an Ensemble Classifier (RF + LR), and two enhanced models Enhanced Logistic Regression and Enhanced Random Forest. Performance metrics such as Accuracy, Precision, Recall, F1 Score, and ROC AUC were computed for comparative analysis. Results showed that the Ensemble (RF + LR) model achieved the highest accuracy (0.8867) and precision (0.7805), followed closely by the Enhanced Logistic Regression with accuracy 0.8861, precision 0.7768, and the highest ROC AUC (0.9740). The Logistic Regression model demonstrated strong overall balance, attaining 0.8860 accuracy, 0.7641 precision, 0.6216 recall, and 0.9755 ROC AUC. Meanwhile, Random Forest achieved stable results with accuracy 0.8854 and ROC AUC 0.9613, whereas KNN recorded the lowest accuracy (0.7887) and weakest F1-score (0.4580), indicating sensitivity to high-dimensional data. The findings confirm that ensemble and optimized models outperform single classifiers in predicting academic performance, with Logistic Regression and its enhanced version showing excellent trade-offs between precision, recall, and AUC. This study underscores the significance of model enhancement and ensemble learning is improving predictive reliability and offers a scalable, interpretable framework for instructors and administrative staff to assist targeted academic interventions and improve student success.