Predictive Modeling for Heart Rate: A Comparative Analysis of LSTM, XGBoost, and LightGBM

This study evaluates the effectiveness of Long Short-Term Memory (LSTM) networks, XGBoost, and LightGBM in predicting heart rates across diverse time windows. Leveraging a dataset from 22 users across multiple observation windows, this research significantly broadens the current understanding of heart rate prediction in the field of personalized health monitoring and sports science. Performance metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Scatter Index (SI) were employed for assessment. Over intervals of 30, 60, and 180 seconds, both XGBoost and LightGBM significantly outperformed LSTM in terms of MAE, MSE, RMSE, and SI. These results suggest that XGBoost and LightGBM are superior options for high-accuracy heart rate prediction, underscoring their potential utility in healthcare and athletic performance monitoring applications.