<![CDATA[Malnutrition among toddlers remains a serious public health issue in Indonesia, with a stunting prevalence of 21.6% in 2022—still above the WHO standard, which sets the maximum threshold at 20%. Traditional methods for assessing nutritional status are time-consuming and prone to human error, highlighting the need for a more efficient and accurate approach. This study aims to develop a system for predicting toddler nutritional status using the XGBoost algorithm, implemented in a web-based application utilizing anthropometric data. A quantitative approach with applied research methods was used, analyzing 5,489 anthropometric records of children from RSUD DR. Gondo Suwarno during the 2017–2023 period, selected through purposive sampling. The dataset included parameters such as age, sex, height, weight, arm circumference, and head circumference of children aged 0–59 months. After data cleaning, 5,169 high-quality samples were retained and divided into 80% training and 20% testing sets with balanced class distribution. The XGBoost model was optimized using Grid Search with 3-fold cross-validation to achieve the best hyperparameter configuration. Results showed that the XGBoost model achieved an accuracy of 97.17%, precision of 97.16%, recall of 97.17%, and F1-score of 97.16% in classifying three nutritional status categories: Normal, Overnutrition, and Undernutrition. Feature importance analysis revealed that weight was the strongest predictor, contributing 42.52%, followed by age (16.79%) and height (15.49%). The system was successfully implemented in a user-friendly web application that allows the input of anthropometric data and provides real-time prediction results. This research produced an effective screening tool for early detection of toddler malnutrition, improving healthcare service efficiency and supporting government programs aimed at reducing stunting rates.]]>
