Construction of a Machine Learning Model to Estimate Physiological Variables of Speed Skating Athletes Under Hypoxic...

Han, J, Liu, M, Shi, J, and Li, Y. Construction of a machine learning model to estimate physiological variables of speed skating athletes under hypoxic training conditions. J Strength Cond Res 37(7): 1543-1550, 2023-Monitoring changes in athletes' physiological variables is essential to create a safe and effective hypoxic training plan for speed skating athletes. This research aims to develop a machine learning estimation model to estimate physiological variables of athletes under hypoxic training conditions based on their physiological measurements collected at sea level. The research team recruited 64 professional speed skating athletes to participate in a 10-week training program, including 3 weeks of sea-level training, followed by 4 weeks of hypoxic training and then a 3-week sea-level recovery period. We measured several physiological variables that could reflect the athletes' oxygen transport capacity in the first 7 weeks, including red blood cell (RBC) count and hemoglobin (Hb) concentration. The physiological variables were measured once a week and then modeled as a mathematical model to estimate measurements' changes using the maximum likelihood method. The mathematical model was then used to construct a machine learning model. Furthermore, the original data (measured once per week) were used to construct a polynomial model using curve fitting. We calculated and compared the mean absolute error between estimated values of the 2 models and measured values. Our results show that the machine learning model estimated RBC count and Hb concentration accurately. The errors of the estimated values were within 5% of the measured values. Compared with the curve fitting polynomial model, the accuracy of the machine learning model in estimating hypoxic training's physiological variables is higher. This study successfully constructed a machine learning model that used physiological variables measured at the sea level to estimate the physiological variables during hypoxic training.