Obesity Prediction Using Synthetic Minority Oversampling Technique for Numeric and Continous and XGBoost Approaches

Abstract

This study investigates the effect of using SMOTE-NC on the XGBoost algorithm in predicting obesity. The main objective of this research is to determine the effect of implementing SMOTE-NC and also the features that are most influential in the prediction process. By using the SMOTE-NC approach with XGBoost it is hoped that it can improve obesity prediction performance, data is collected from UCI Machine Learning for Obesity analysis. The prediction results reveal that the application of SMOTE-NC can improve the accuracy of obesity prediction using XGBoost. The results show that the best accuracy in this study was able to reach 98.30%. Further analysis, this research reviews several influential features in the prediction process, namely Weight, Height and Age. Based on these results, it is hoped that they can contribute to further research. Overall, this research underlines the importance of maintaining health to avoid obesity by keeping body weight within normal limits.