A Deep Learning Approach to Calorie Expenditure Prediction: Including Exercise Intensity and Environmental Factors

Calorie expenditure prediction must be accurate for building physical fitness and personal health management. Many studies have focused on the essence of machine learning approaches to predict calorie burn using existing studies XGBoost, linear regression, and random forests, but they totally ignore some of the most important factors, such as exercise intensity and weather, which are significant contributors to how many calories we burn. Herein, we introduce a Deep Neural Network (DNN)-based framework for predicting calorie burnsthat uses the unexplored features described above. We consider a dataset from Kaggle that consists of features such as heart rate, BMI, duration of activity, exercise intensity, and weather conditionsin place to illustrate that our DNN outperforms classical models (lower Mean Absolute Error: 0.86 vs. the baseline XGBoost model MAE = 1.48). Advanced activation functions and optimization techniques were used for training in fewer epochs with greater efficiency. An ablation study revealed that exercise intensity and weather conditions have a major impact on prediction accuracy. Our findings provide potential to enhance fitness analytics as a reliable device for personalized health coaching and wellness monitoring. This work advances our understanding of machine learning-based approaches to calorie prediction in health applications and provides a basis for future studies on dynamic calorie burn prediction