A Context-Aware Machine-Learning Early-Warning Model for O-Level Dropout Risk in Southwestern Uganda

Uganda's lower-secondary completion rate stood at just 26.4% as of UNESCO's most recent data cited in UNICEF's 2024 policy brief on the country's education sector, with the Ministry of Education and Sports attributing over 40% of dropouts to financial hardship.
Researchers at Bishop Stuart University built a context-aware machine-learning model to flag O-Level students at risk of dropping out in six private mixed secondary schools across Mbarara and Isingiro districts, drawing on complete records for 331 of 1,919 accessible learners. Comparing logistic regression, random forest, support vector machine and k-nearest-neighbour classifiers with five-fold cross-validation, the team found SVM performed best (F1-score 0.633), ahead of logistic regression (0.600), random forest (0.511) and KNN (0.468). Chronic absenteeism, financial hardship and delayed fees, long travel distances, weak examination confidence, bullying, limited parental engagement and psychological stress were the leading risk indicators.
The authors are candid about the model's limits: a 'moderate F1-score, small purposive school frame, researcher-constructed risk labels and absence of external prospective validation preclude autonomous high-stakes use.' They recommend it only as a human-supervised screening aid feeding into counselling and academic-support referrals, not as a stand-alone decision tool — an honest caveat for any school considering similar systems.

Read the full analysis: https://journals.eanso.org/index.php/eajis/article/view/5818