<![CDATA[Stunting is a serious public health problem, particularly in Balangan Regency, where the prevalence in 2022 reached 29.8%, exceeding the national average of 21.6%. This condition highlights the importance of data-driven early detection to support targeted prevention efforts. This study aimed to develop an early stunting risk classification model using historical toddler healthcare data and evaluate the performance of several Machine Learning classification algorithms. Unlike previous studies that generally focused on a limited number of classification methods, this study provides a comprehensive comparison of eight Machine Learning algorithms, namely Naive Bayes, K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Random Forest, AdaBoost, XGBoost, LightGBM, and CatBoost, combined with Synthetic Minority Oversampling Technique (SMOTE) and Recursive Feature Elimination (RFE) to improve classification performance. The dataset was divided into 80% training data and 20% testing data. Model evaluation was conducted using 5-fold cross-validation with accuracy, precision, recall, and F1-score metrics. The results showed that AdaBoost achieved the best performance with an accuracy of 99.39%, precision of 100%, recall of 98.78%, and F1-score of 99.39% during cross-validation. Further evaluation on testing data produced an accuracy of 90.20%, precision of 90.57%, recall of 92.36%, and F1-score of 91.39%. The selected model was implemented into a Streamlit-based web application to support both manual and bulk prediction. The developed system can assist healthcare institutions and healthcare workers in conducting accurate and efficient early stunting risk detection, enabling more targeted interventions and data-driven decision-making. Therefore, this study provides both methodological and practical contributions to support stunting prevention efforts.]]>
Copyrights © 2026
