Predicting School Dropout in Malawi

School dropout is a significant issue, especially in developing countries due to high poverty levels and inadequate allocation of resources to education. This study applies Machine Learning to predict dropouts in the Lilongwe University of Agriculture and Natural Resources’ open and distance education system. Four supervised machine learning classifiers (Gaussian Naïve Bayes, Logistic Regression, K-Nearest Neighbour, and Random Forest) were assessed to find the best predictor for dropouts. Data imbalance was addressed using oversampling and undersampling techniques. Results showed that Random Forest performed the best with under-sampling. Hyperparameter optimization using grid and random search methods also improved performance, with Random Forest emerging as the best classifier. This study contributes to future research and enhances the existing literature. It is expected to improve student support services by proactively addressing at-risk students, reducing attrition rates.