<![CDATA[Stunting and nutritional issues among children under five remain major public health challenges in Indonesia, requiring precise, data-driven monitoring. This study aims to establish nutritional status profiles for children under five using a machine learning approach based on anthropometric indicators. The methodology began with a data quality audit and Exploratory Data Analysis (EDA) on a dataset comprising 120,999 observations with no missing values. The dataset consisting of age, height, and gender attributes was split using a stratified approach (80:20) and subsequently standardized using StandardScaler and OneHotEncoder. Four classification algorithms (Logistic Regression, Decision Tree, Gradient Boosting, and Random Forest) were evaluated comparatively. Experimental results demonstrated that Random Forest achieved the best performance, with an accuracy of 98.94%, a macro-averaged Precision of 98.74%, a macro-averaged Recall of 98.45%, and a macro-averaged F1-Score of 98.59%. The model's reliability and consistency were confirmed via 5-Fold Stratified Cross-Validation, yielding an average accuracy of 99.04% with a very low standard deviation (0.0009). Furthermore, feature importance analysis revealed that height (60%) and age (39%) were the most dominant anthropometric indicators in determining the nutritional status profiles of children under five. This study demonstrates that the Random Forest algorithm, based on anthropometric indicators, is suitable for implementation as a decision support system for growth monitoring and the early detection of nutritional issues in children.]]>
Copyrights © 2026
