Student Success Through Algorithms: An Analytical Study of Machine Learning Models

Accurate prediction of student academic performance is essential for early interventions and improved educational outcomes. This research explores machine learning (ML) approaches to forecast student success using educational data mining (EDM) techniques. Leveraging a dataset from the University of California Irvine with 4,424 student records and 35 features, the study applies ensemble learning methods—including Decision Trees, Random Forests, XGBoost, and CatBoost—alongside Naïve Bayes classifiers to predict three academic outcomes: graduation, dropout, and continued enrollment. The analysis incorporates exploratory data analysis, feature selection using chi-squared testing, and class imbalance correction through SMOTE. Performance metrics such as accuracy, F1 scores, and confusion matrices indicate that CatBoost outperforms other models, achieving an accuracy of 76% and a dropout prediction F1 score of 0.87. The findings provide actionable insights for academic institutions to identify at-risk students and implement targeted support strategies. The study concludes with a discussion of model limitations and directions for future research in predictive analytics in education.