Gradient Boosting Machine Algorithm Optimization for Child Nutrition Classification at Al-Islam Hospital, Krian

Abstract

The double burden of malnutrition in school-aged children (6–12 years old) remains a serious challenge for the healthcare sector in Indonesia. Conventional manual monitoring methods currently implemented in hospitals often face time constraints and are susceptible to human error. This study aims to optimize the architecture and parameters of the Gradient Boosting Machine (GBM) algorithm in classifying children's nutritional status digitally, quickly, and precisely at Al-Islam H.M. Mawardi Krian Hospital. The research dataset includes 257 medical records of pediatric patients from 2023–2025 containing anthropometric features such as age, gender, weight, height, and Body Mass Index (BMI). Data pre-processing stages are carried out through gender encoding, feature engineering, target binary transformation ("Good" and "Not Good"), feature selection, minority data balancing with SMOTE, and value standardization. The GBM model was configured using a parameter combination of 200 trees (n_estimators), a learning rate of 0.2, a maximum depth of 4 levels, and a subsample of 0.8. Test results demonstrated impressive and stable model performance, with an accuracy rate of 92% (0.92). In the 'Good' nutritional status category, the model achieved a precision of 0.94 and a recall of 0.98, while in the 'Poor' category, it achieved a precision of 0.80 and a recall of 0.57. This study concluded that the application of the boosting-based ensemble method is highly effective in improving the accuracy of clinical classification of children's nutritional status and has the potential to be integrated into hospital information systems as a real-time decision support system.