Predicting the Neonatal Mortality Using Ensemble Machine Learning Algorithms in Case of Ethiopian Rural Areas

Abstract

Each year, approximately 2.5 million newborns die globally, with developing countries behavior the impact of this crisis. Sub-Saharan Africa experiences the highest neonatal mortality rate at 27 deaths per 1,000 live births. In Ethiopia, neonatal mortality remains alarmingly high at 29 deaths per 1,000 live births, with early neonatal mortality reaching 41.8 deaths per 1,000 live births. Rural areas face even more severe disparities, with a prevalence of 45.6 deaths per 1,000 live births compared to 25.5 in urban settings, basically due to inadequate healthcare access, poor maternal and neonatal services, and socioeconomic challenges. This study aimed to develop a robust predictive model for neonatal mortality in rural Ethiopia, using secondary data from the Ethiopian Demographic and Health Surveys (2000–2019). The dataset, consisting of 29,048 instances and 22 relevant features, was preprocessed to handle missing values and balance the class distribution using SMOTE. Several advanced ensemble machine learning algorithms were applied to build the predictive model, including Random Forest, Gradient Boosting, Extreme Gradient Boosting, Light Gradient Boosting, and CatBoost. The performance of these models was evaluated based on key metrics, including accuracy, precision, recall, F1 score, and ROC-AUC. Among the ensemble algorithms tested, CatBoost demonstrated the highest performance, achieving 97.5% accuracy, 97.52% precision, 97.5% recall, 97.5% F1 score, and an outstanding ROC-AUC value of 99.57%. The key risk factors for neonatal mortality identified in the study included BCG vaccination status, the number of under-five children in the household, recent episodes of diarrhea, and iron tablet intake during pregnancy. These factors were found to significantly contribute to predicting neonatal mortality, underscoring the importance of targeted healthcare interventions for high-risk neonates. This study developed a predictive model for neonatal mortality in rural Ethiopia using ensemble machine learning, identifying key risk factors like BCG vaccination and maternal health. It offers actionable insights for targeted interventions, supports healthcare prioritization, and highlights the need for improved access and policy reforms. Mobile health apps and policymaker collaboration can further reduce neonatal mortality.