Calories Burnt Prediction Using ANN and XGBoost Regression Algorithms from Physical Activities

Predicting the number of calories burned during physical activity is essential for managing personal fitness, health monitoring, and preventing lifestyle diseases. This study offers a reliable and hybrid method for precisely estimating calories burned from physical activity data by merging Artificial Neural Networks (ANN) and the XGBoost regression algorithm. The model uses attributes including heart rate, duration, activity type, and body characteristics to improve forecast accuracy. While the gradient-boosting framework XGBoost improves generalization and addresses overfitting, the ANN component captures intricate, non-linear correlations between input characteristics. By combining these advantages, the suggested methodology outperforms independent models in terms of prediction. Results show that the hybrid model performs better than conventional regression models in terms of mean absolute error (MAE) and root mean square error (RMSE) while training and validating on a dataset of physical activity logs. The study also emphasizes how crucial feature engineering and hyperparameter adjustment are to attaining the best accuracy. With potential uses in wearable technology and healthcare platforms, the research's conclusions promote data-driven solutions for fitness tracking and tailored health management.