Performance Comparison of Logistic Regression and Neural Networks in Predicting Stunting Prevalence in Aceh Province

Abstract

Stunting remains a significant public health problem in Indonesia, particularly in Aceh Province, where differences in prevalence across districts/cities require data-driven approaches to support targeted intervention planning. This study aims to compare the predictive performance of Logistic Regression and Neural Networks in classifying stunting prevalence across 23 districts/cities in Aceh Province. A quantitative approach was employed using secondary data from the Aceh Provincial Health Profile for 2022–2025, comprising 92 observations. The predictor variables included exclusive breastfeeding coverage, improved sanitation coverage, and vitamin A supplementation coverage, while stunting prevalence was transformed into a binary classification, with 0 representing prevalence of ≤20% and 1 representing prevalence of >20%. Logistic Regression and Multilayer Perceptron Neural Networks were implemented and evaluated using Accuracy, Precision, Recall, F1-score, and Area Under the Curve (AUC). The results showed that both models were capable of classifying stunting prevalence, while Neural Networks demonstrated better overall predictive performance. Logistic Regression achieved an Accuracy of 0.78 and an AUC of 0.82, whereas Neural Networks achieved an Accuracy of 0.89, Precision of 0.87, Recall of 0.90, and F1-score of 0.88. The findings indicate that Neural Networks provide stronger predictive performance in identifying high-stunting categories, while Logistic Regression offers greater interpretability for understanding relationships between health indicators and stunting prevalence. Therefore, Neural Networks have potential as a predictive approach for supporting the identification of high-risk areas and data-driven stunting intervention planning in Aceh Province.