Calorie Predictive Activity Analyzer using ML

Abstract

The aim of this study is to investigate the possibility of using machine learning methods to estimate calorie expenses during training. The dataset used to train machine learning models included 15,000 recording with seven main variables, such as heart rate, temperature and activity. These models were XGBOOST, linear regression, random forest and support vector machine (SVM). With an absolute error of about 0.94 calories (MAE), XGBOOST improved others, 99.67% received on both training and test dataset. The conclusions have significant effects to create smart, personal fitness tracking systems, as they show that machine learning is good for mimicking physical energy consumption. Thanks to the extra support for the inclusion of such models in health monitoring apps and weekly tools, smart and more adaptive welfare solutions on the horizon.