Analisis Prediktif Faktor Kematian Balita di Bandung menggunakan Logistic Regression, Random Forest, dan XGBoost

Kharismawardani, Aqila
Purnama, Denny Ganjar

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 6 (2025): November 2025

The Under-Five Mortality Rate (UFMR) is a crucial issue in Indonesia that requires data-driven interventions. This study aims to develop a predictive model to identify the most influential risk factors for under-five mortality in Bandung City and to compare the performance of three machine learning algorithms. This research utilizes secondary data from the Bandung City Open Data portal for the period 2019-2021. The method employed is a comparative analysis of Logistic Regression, Random Forest, and XGBoost. To address the significant class imbalance in the data, the Synthetic Minority Over-sampling Technique (SMOTE) was applied to the training data. The evaluation results show that all three models achieve high accuracy, however, performance on the minority calss (mortality cases) remains challenging, indicated by low F1-scores (0.12 for Random Forest and 0.17 for XGBoost). Nonetheless, the feature importance analysis from the Random Forest model successfully identified 'other causes' (penyebab_LAIN-LAIN), 'fever' (penyebab_DEMAM), and the availability of healthcare professionals (PERAWAT, BIDAN) as the most significant predictors. This study highlights the insight from feature importance in identifying risk factors in imbalanced medical data, providing a basis for more targeted health policy recommendations.