Rédiger un PREreview

In India and across the world, obesity is a serious public health issue, creating an urgent need for accurate predictive models. This research analyzes machine learning algorithms to predict obesity within a statistically optimized framework. The proposed approach compares classical and advanced models: Logistic regression, decision tree, random forest, k-nn, naive bayes, LGBM Classifier, XGBoost Classifier, Gradient Boosting, and DNN.The dataset contains factors such as body height, body weight, dietary behaviour, and physical activities. The preprocessing phase entailed features selection, normaliza-tion, and handling of missing data. Efficiency was calculated using accuracy, recall, precision, F1-score, and AUC.LGBM performed much better than the rest of the algo-rithms with an accuracy rate of 95.74%, followed by XGBoost (95.27%), gradi-entBoost(95.09%), DNN (94.79%), Decision Tree (94.08%), k-nn (91.48%), and naive bayes (82.50%). Accordingtothese findings, tree-based models performed better in forecasting obesity compared to traditional BMI models and can be applied instead of them. This study emphasizes the significance of taking into consideration not only physical activi-ties but also nutritional behaviours while trying to solve the problem of obesity.