Revolutionizing Workout Analytics: Machine Learning Models for Calorie Burn Estimation

Abstract

In predicting calories burned based on personal data, researchers aimed at deploying machine learning. The first step involved data preparation and cleaning, which included imputation of missing values, before undertaking Exploratory Data Analysis. Five different machine learning models were used: K-Nearest Neighbors (KNN), Decision Tree (DT), AdaBoost (AB), Support Vector Machine (SVM) and XGBoost (XGB). Both default and optimized hyperparameters were utilized to train the models for better outcomes. Some of the models predicted calories burned successfully with XGBoost being the best fit. Feature interactions influencing the output became apparent through Explainable AI which was employed for this purpose. This study emphasized on hyperparameter tuning as a way to achieve optimal results when using any given model. With an RMSE value of 2.13, and a coefficient of determination R² equal to one, XGBoost carded home the best performance. This research shows that it is possible to apply machine learning in order to forecast calorie loss from individual data, hence improving fitness and health programs.