Machine Learning-Based Prediction of Performance Gaps in Rowing and Identification of Key Training Monitoring...

Area of Science:

• Sports Science
• Biomechanics
• Data Analytics

Background:

• Elite rowing relies on precise biomechanical data from sensors.
• Systematic integration of sensor data into unified datasets is lacking for practical decision support.
• Translating multidimensional performance indicators into actionable insights remains a challenge.

Purpose of the Study:

• To construct a unified rowing training monitoring dataset from sensor-derived biomechanical measurements.
• To develop predictive models for athletes' performance gaps relative to target 2 km performance.
• To identify key training monitoring indicators for target attainment classification and evaluate their practical value.

Main Methods:

• Utilized 249 biomechanical testing records from the Chinese National Rowing Team (2024-2025 season).
• Generated 449 athlete-level records after standardized processing, with 172 for modeling.
• Employed XGBoost Regressor, Elastic Net, and LASSO for regression, and Logistic Regression and XGBoost Classifier for classification tasks, with athlete-level cross-validation.

Main Results:

• All models outperformed baseline models; XGBoost Regressor excelled in MAE, while Elastic Net led in RMSE and R².
• Key predictive indicators included mean boat velocity, minimum boat velocity, stroke rate, distance per stroke, and efficiency metrics.
• XGBoost Classifier achieved an ROC AUC of 0.992 for target attainment classification, with stable performance independent of grouping variables.

Conclusions:

• Developed an applied framework for integrating sensor data in elite rowing for performance evaluation and decision support.
• Identified key biomechanical and technical indicators crucial for predicting rowing performance.
• The study highlights the potential of data-driven approaches to optimize training in elite sports settings.