Random Forest Classification of Infant Mortality Rate in Indonesia: A Gini-Based Analysis

One of the indicators used to measure the success of development programs in Indonesia is the Infant Mortality Rate (IMR). IMR is a sensitive indicator and represents maternal and child health problems in a country. Random forest is an ensemble machine learning method that combines multiple decision trees using bootstrap aggregation. It aims to improve the prediction accuracy and robustness of the model. In addition, it can be applied to both case classification and regression because it can handle high-dimensional and complex cases and non-linear relationships. In this study, Random Forest is used to solve the classification of IMR cases in Indonesia, making them easy to interpret and related to policy relevance. The aim of this study is to predict infant mortality factors using the Gini Index to determine which variables need to be improved. The Gini Index is used to identify key factors, enabling targeted policy interventions. It highlights the most influential variables, helping policymakers focus on areas that require improvement for more effective outcomes. The evaluation model in this study uses out-of-bag estimation and k-fold validation. The model achieves an overall accuracy of 99.97%, with a sensitivity of 99.87% and specificity of 100\%, indicating excellent performance. The most important variables in this study are breastfeeding, type of birth (single and twin), and birth weight of the baby. The parent node in IMR is breastfeeding, where live IMRs that are breastfed have a greater chance of survival than dead IMRs that are not breastfed.