Comparative Analysis of Machine Learning Models for Stunting Prediction in Jakarta

Abstract

Stunting is one medical problem that inhibits a baby's growth. Prompt diagnosis is essential to prevent long-term harm. This study compares machine learning techniques, including Naïve Bayes, Decision Tree, Random Forest, SVM, and ensemble methodologies, in order to improve prediction accuracy. Information on 1,723 children in Jakarta, including age, height, gender, family health history, household income, access to health services, and hygienic circumstances, is included in this dataset, which was collected from Riskesdas and hospital and clinic medical records. To improve model performance, SMOTE, feature selection, and normalization techniques were used. The ensemble approach combined Naïve Bayes with Decision Trees via stacking. The assessment findings indicated that Random Forest had the best accuracy (98%), followed by ensemble technique and Decision Tree (97%), while Naïve Bayes and SVM had lesser accuracy (38% and 37%). This model can assist the government in early intervention to prevent stunting.