Comparing Gradient Boosting and Linear Models for Calorie Prediction

This study presents result of experiments of five models like Linear Regression Model, Ridge Regression Model, Lasso Regression Model, Random Forest Regressor, and XGBoost Regressor on prediction of calorie expenditure by exercising biology of data. The models were evaluated with respect to Mean Absolute Error (MAE), Mean Squared Error (MSE), and R2 scores. Results imply that ensemble models, particularly XGBoost Regressor, performed much better than traditional linear models. XGBoost achieved the best performance with the least MAE (12.56) and MSE (191.63), together with highest <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2}$</tex> value of (0.96) followed in performance by Random Forest Regressor. Linear models, Linear Regression, Ridge, and Lasso, showed high error and low <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2}$</tex> scores, indicating their limitation in taking into account the non-trivial non-linear relationships embedded in the data. These findings underscore the important of appropriate choice of machine learning techniques for accurate calories prediction and suggested ways in which linearity gave way for shallower examination by feature expansion and model optimization.