Comparison of Classification Results of Categorical Boosting and Extreme Gradient Boosting Methods in Obesity Class Diagnosis
P Jonatan, Giska Auria, Anita Desiani, Ali Amran, Indri Ramayanti, Irmeilyana

Universitas Sriwijaya

Abstract

Obesity, caused by a chronic energy imbalance in which calorie intake exceeds expenditure, is a growing global health problem that negatively impacts quality of life. Early detection is crucial for mitigation. This study compares the performance of the Extreme Gradient Boosting (XGBoost) and Categorical Boosting (CatBoost) algorithms for obesity status classification. The research methodology included data collection, pre-processing, and model evaluation using two validation strategies: percentage split and k-fold cross-validation. The results show the superiority of CatBoost. Using a percentage split, CatBoost achieved an accuracy of 92.53% compared to XGBoost (91.67%). With k-fold cross-validation, CatBoost (91.95%) also outperformed XGBoost (90.46%). Additionally, CatBoost consistently produced superior average values for precision, recall, and F1-score. It is concluded that CatBoost provides a more accurate and robust model for obesity classification compared to XGBoost. Future research could explore feature engineering techniques or other algorithms to further enhance predictive accuracy.n Edit It Again Later

Keywords: Boosting Algorithm- CatBoost- Classification- Machine Learning- XGBoost

Topic: Mathematics and Applications