Predictive modeling of energy expenditure during treadmill exercise: An integrated approach combining body composition measurements and machine learning

Abstract

Accurately estimating caloric expenditure during physical activity remains a significant challenge in sports science،rehabilitation medicine،and personalized fitness programs. To develop an effective predictive model for caloric expenditure during treadmill exercise،this study proposes an integrated approach that combines detailed body composition analysis with Artificial Intelligence (AI) techniques. Data were collected from 51 participants and included comprehensive body composition measurements،such as body fat percentage،muscle mass،body water content،protein percentage،Basal Metabolic Rate (BMR)،Body Mass Index (BMI)،skeletal muscle ratio،visceral fat index،biological age،and waist-to-hip ratio. In addition،caloric expenditure data were recorded at ten treadmill speed levels ranging from 3 to 12 km/h. A Machine Learning (ML) model was developed and trained to predict caloric expenditure based on body composition characteristics. Feature importance analysis using SHAP (Shapley Additive Explanations) values identified BMR، muscle mass ،and body fat percentage as the most influential predictors. Model performance was evaluated using several statistical metrics ،including the coefficient of determination (R²)،Root Mean Square Error (RMSE)،and Mean Absolute Error (MAE). The results demonstrated that the AI-based model outperformed conventional regression approaches in predicting caloric expenditure. The findings highlight the potential of combining body composition analysis with AI techniques to provide accurate، individualized estimates of caloric burn through body scanning technologies.