Abstract
Measles remains a significant public health concern, particularly in low- and middle-income countries, where vaccination uptake is often suboptimal. In 2021, an estimated 128,000 measles deaths were reported globally, mostly among unvaccinated or under-vaccinated children under five years. Machine learning (ML) presents tremendous opportunities to predict measles vaccination status and support the development of targeted interventions, but their implementation is still nascent. Therefore, this study applied machine learning models to predict measles vaccination status among children under five in Zimbabwe using secondary data from the 2019 Zimbabwe MICS. Label encoding was used to preprocess categorical variables and the Synthetic Minority Over-sampling Technique (SMOTE) to address the class imbalance. Three ensemble ML classifiers—AdaBoost, Random Forest and Bagging were evaluated using accuracy, precision, recall, F1-score and AUC-ROC. A total of 334 observations were analyzed in this study and about 83% of children under five years received measles vaccination. The statistical analysis revealed that the age of the child (p = 0.000) and wealth index (p = 0.015) were significant predictors of measles vaccination uptake, with older children and those from wealthier households being more likely to be vaccinated. Other factors such as the sex of the child, region, place of residence, mother’s education level and health insurance coverage were not statistically significant. Random forest classifier achieved the highest AUC-ROC (0.96), followed by AdaBoost (0.95) and Bagging (0.94). AdaBoost recorded the highest recall (0.949), making it effective at identifying vaccinated children, while the Bagging classifier achieved the highest precision (0.868), minimizing false positives. The findings suggest that public health efforts should prioritize younger children and low-income households to improve vaccination rates.