Predicting the Speed of Chinese Elite Cross-Country Skiers: A Neural Network Approach Based on the Measurement of Key Biomechanical and Physiological Parameters

This study aimed to identify key biomechanical and physiological parameters affecting cross-country skiing performance and develop a neural network model for predicting skiing speed. Biomechanical attributes (cycle length and rate, vertical displacement of the center of mass, and angular kinematics) and physiological factors (maximal oxygen uptake, 30 s anaerobic power), along with physical fitness (standing long jump, pull-ups) were assessed for 82 cross-country skiers (52 men and 30 women). Random forest analysis was utilized to identify the most influential parameters on skiing speed, which were subsequently used as input parameters to develop a neural network aimed at predicting this speed. The findings identified the primary predictors of skiing speed as the cycle length on both flat and uphill terrains, vertical displacement of the center of mass during the poling phase on uphill terrain, maximal oxygen uptake, and 30 s anaerobic power. The developed neural network model demonstrated high precision in predicting skiing speeds, evidenced by a strong correlation with actual speeds (correlation coefficient of 0.953) and 97.1% of predictions falling within the 95% Bland–Altman agreement limits, affirming the model’s reliability and effectiveness in forecasting skiing performance.