Optimizing Gradient Boosting Algorithms for Obesity Risk Prediction: A Comparative Analysis of XGBoost, LightGBM, and CatBoost Models

The importance of should be given to the detection of obesity risk. In the context of this study, the comparison of three machine learning models to the detection of obesity risk demonstrates that the most accurate model is an XGBoost classifier with an accuracy of 91%, while LightGBM and CatBoost models were slightly less accurate, with the accuracy of 90%. These models may become significant tools in recognizing people at risk of obesity to promote preventive and personalized healthcare that could address the alarming rates of obesity-related complications. Since the models drive on extensive datasets of various demographic and clinical characteristics, including the following features and variable importance plot, Weight 0.38, Height 0.11, Age 0.10, Frequency of Consumption of Vegetables FCVC 0.098, Gender 0.06, Time Using Technology TUE 0.045, Consumption of Water Daily CH2O 0.045, Physical Activity Frequency FAF 0.036, Number of main meals NCP 0.031, Family history with overweight 0.025; CAEC 0.023; CALC 0.021; Mode of transport MTRANS 0.013; Frequency of Consumption of High Caloric Food FAVC 0.011; Type of Caloric Beverage SCC 0.005, they have the capacity to identify subtle patterns and trends that may indicate obesity risk. These pieces of knowledge can empower practitioners to implement individualized approaches to obesity prevention and control strategies. The accurate identification of people at risk of obesity may be a breakthrough discovery to promote public health. Person-time detection can help identify at-risk people for targeted preventive efforts, starting with lifestyle modification and patient education and extending to pharmacological and technological means.