The purpose of this study was to train numerous predictive models to investigate if resting heart rate variability (HRV) can predict elite track and field sprint performance. Dataset encompassed fifteen male Division I NCAA track and field student-athletes who monitored HRV 24 hours prior to competition throughout their competitive careers, accruing 182 total samples. Ensemble machine learning techniques (bagging, boosting, and stacking) were trained from athlete`s HRV to forecast their individual competition performance. Bagging random forest algorithm resulted in mean accuracy of 75.0 ± 8.0%, stochastic gradient boosting achieved 81.2 ± 6.8% mean accuracy, and stacking meta-classifier achieved mean accuracy of 85.6 ± 4.8%. DC Potential, RMSSD, and SDSD were the most predictive readiness indicators towards sprint performance. Estimated functional relationships suggest DC Potential between 25-35 mV and log-transformed RMSSD around 4.5 ms indicates a physiological status that may promote optimal sprint performance
© Copyright 2018 OA Journal - Sports. EnPress. Alle Rechte vorbehalten.