Exploration of Heart Rate Variability for the Prediction of Performance in Youth Footballers

This study aimed to evaluate the prediction quality of short-term resting heart rate variability (HRV) parameters for estimating endurance performance in youth footballers.Male players (n = 222, mean age 13 2.3 years) underwent 5-minute RR intervals (RRi) measurement in a supine position using a heart rate monitor.RRi stationarity was confirmed using the Augmented Dickey-Fuller test ( = 0.05).Time-domain and nonlinear HRV parameters were calculated.The data was split 75:25 for training and testing.Gradient Boosting (GBR), AdaBoost (ABR), and Random Forest (RFR) were used to predict VO2max from the Yo-Yo Intermittent Recovery Test.Models were validated using mean absolute error (MAE) and root mean square error (RMSE).Key HRV parameters were ranked using explainable artificial intelligence (XAI) tools.78% of RRi were stationary.The GBR model achieved MAE: 4.2, RMSE: 5.1; ABR achieved MAE: 3.7, RMSE: 4.4; and RFR achieved MAE: 3.8, RMSE: 4.6.Key features included mean RRi, DFA2, and SD1/SD2.Machine learning models was used to evaluate the quality of endurance performance prediction.They enabled to identify significant HRV parameters that can be further analyzed in this regard.