Hybrid Model Analysis for Calorie Prediction Using Ensemble Learning Techniques: XGBoost and Random Forest

Abstract

The accurate estimation of calorie expenditure during physical activity is critical for personalized fitness management and health assessment. This study uses advanced ensemble learning methods (XGBoost and Random Forest regressors) to estimate calories burned based on demographic and physiological variables such as gender, age, height, weight, exercise duration, heart rate, and body temperature. The dataset was preprocessed to ensure quality and consistency prior to model training. Performance was assessed using R², RMSE, and MAE. Comparative results showed that XGBoost had better predictive performance and generalization due to its gradient boosting framework and built-in regularization. Streamlit was used to create a web-based interface that allows end users to estimate their calories in real time. The findings highlight the potential of ensemble methods in fitness data analysis and provide a scalable solution for integrating calorie prediction into digital health platforms. Future extensions may include diverse exercise categories and real-time data acquisition from wearable sensors to enhance prediction accuracy.