Statistical Insights Into Machine Learning Models for Predicting Under-Five Mortality: An Analysis From Multiple Indicator Cluster Survey (MICS)

Under-five mortality remains a significant public health challenge in the Central African Republic (CAR), yet research on the region-specific determinants of this issue is limited. This study explores the use of machine learning (ML) models to predict under-five mortality in CAR using data from the 2018-2019 Multiple Indicator Cluster Survey (MICS). Advanced ML models, including CatBoost, Light Gradient Boosting (LGB), Extreme Gradient Boosting (XGBoost), and Extreme Gradient Boosting Random Forest (XGBRF), were compared with traditional logistic regression, with model performance evaluated using accuracy metrics such as the Area Under the Curve (AUC), F1 score, Precision, Recall, and Matthews Correlation Coefficient (MCC). CatBoost outperformed all models, achieving an AUC of 0.973, accuracy of 0.946, F1 score of 0.951, and MCC of 0.897, demonstrating its robust predictive power. The study identified key factors influencing under-five mortality in CAR, such as birth type, maternal education, birth order, birth interval, and birth weight. It suggests that using advanced ML models, particularly CatBoost, can improve identifying region-specific risk factors and inform targeted interventions to reduce child mortality in CAR.