Identification of Athleticism and Sports Profiles Throughout Machine Learning Applied to Heart Rate Variability.

Area of Science:

• Sports Science
• Biomedical Engineering
• Data Science

Background:

• Heart rate variability (HRV) is a key non-invasive indicator of physiological status.
• Machine learning (ML) offers advanced analytical capabilities for complex HRV datasets.

Purpose of the Study:

• To identify athletic characteristics using HRV and ML algorithms.
• To develop models for classifying athletes versus non-athletes and identifying individual soccer players within a team.

Main Methods:

• Two models were developed: M1 for athlete classification (856 athletes, 494 non-athletes) and M2 for soccer player identification (105 players, 514 teammates).
• Applied machine learning algorithms: Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM).
• Utilized SHAP values for model interpretation.

Main Results:

• SVM achieved the highest performance in M1 (accuracy=0.84, ROC AUC=0.91).
• Random Forest performed best in M2 (accuracy=0.92, ROC AUC=0.94).
• Proposed athleticism and soccer identification indices derived from HRV data.

Conclusions:

• ML algorithms (SVM, RF) can effectively generate HRV-based indices for athletic identification.
• These indices aid in distinguishing athletes and identifying specific sports profiles.
• Systematic integration of HRV assessment into training regimens is recommended for enhanced athletic evaluation.