An Empirical Analysis for Predicting Obesity Levels Using Machine Learning Models

This is the authors' abstract. We don't add key points for this paper.

Obesity is a significant global health issue that has been steadily increasing due to factors like sedentary lifestyles, unhealthy dietary habits, and reduced physical activity. Predicting obesity levels based on lifestyle factors can help in early intervention and targeted prevention strategies. Machine learning could be a good approach used for early detection and intervention of obesity levels in an indi.vidual. In this study, an obesity detection dataset was used to examine key factors such as physical activity, dietary habits, alcohol consumption etc. and how they are contributing to obesity. Several ML algorithms such as XGBoost, Gaussian Naive Bayes (NB), CatBoost, LightGBM, etc. have been used for classification and evaluation metrics such as accuracy, f1-score, precision, recall, Jaccard score, log loss, ROC-AUC score, and average cross-validation score were used to evaluate each model. For Obesity categorization, the LightGBM excelled in all techniques with the highest accuracy score of 97%.