<![CDATA[Child stunting remains a major public health challenge in Indonesia and continues to hinder progress toward the Sustainable Development Goals (SDGs), particularly in child health and nutrition. Early identification of at-risk children is therefore essential to support timely interventions. While previous machine learning studies on stunting prediction commonly incorporate socioeconomic, environmental, and behavioral variables, comparative evaluations based exclusively on routinely collected anthropometric indicators remain limited, particularly within Indonesian primary healthcare settings. This study evaluates the predictive performance of multiple machine learning models for stunting classification using only anthropometric and early-life growth indicators. A dataset consisting of 1,000 child records—including age, birth weight, birth length, current weight, current length, and breastfeeding status—was analyzed using Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, and Gradient Boosting algorithms. The dataset was partitioned using an 80:20 stratified train–test split, while five-fold cross-validation was applied during model development to improve robustness and reproducibility. Experimental results demonstrate that ensemble-based methods outperform single classifiers, with Gradient Boosting achieving the highest predictive performance (accuracy = 0.90, F1-score = 0.90, AUC = 0.93). Feature importance analysis reveals that birth length, birth weight, current weight, and age are among the most influential predictors of stunting risk. These findings suggest that machine learning models built solely on routinely collected anthropometric indicators can provide a practical, scalable, and data-driven approach for early stunting detection in Indonesian primary healthcare systems.]]>
Copyrights © 2026
