Predicting Infant Mortality Using Machine Learning Regression Models: A Case Study of Somalia

In Somalia, mortality among infants persists one of the most urgent health issues, which ranked to be the 4th highest infant mortality rate in the world by 2025. The need for data to improve health care policy and equitable resource allocation in low-income settings motivates this research. The authors of this study use Machine Learning to predict and evaluate several ML regression models to predict monthly infant mortality in Somalia. Different model types were tried, including linear regression, random forest, KNN and SVM. They used baseline data from the Somali health system to identify key factors affecting infant mortality and to build the most effective prediction model applying demographic and vaccination data analysis. A dataset of 2,955 records and 22 columns was used and extensive preprocessing techniques were applied including missing values, outlier handling, feature scaling, and the use of SMOGN while considering data imbalance. The performance of all these were checked; Of these, the Random Forest model had the highest performance in predicting infant mortality, achieving an R2 score of 84.15%, a mean absolute error (MAE) of 2.83, and a mean square error (MSE) of 38.07. As a result of this study encourage data-driven policies to address public health concerns and optimize healthcare costs, thus supporting predictive health analytics research.