Identifying Weekly Physical and Physiological Profiles in Professional Basketball Using Heart Rate Variability and...

Area of Science:

• Sports Science
• Physiology
• Data Science

Background:

• Understanding physiological load is crucial for optimizing athletic performance and preventing injury in professional sports.
• Individual responses to training load vary significantly among athletes, necessitating personalized monitoring strategies.

Purpose of the Study:

• To apply unsupervised machine learning to characterize distinct weekly physiological load states in professional basketball players.
• To identify and describe player profiles based on autonomic, cardiovascular, and mechanical load parameters over an 18-week season.

Main Methods:

• Utilized K-means clustering on weekly standardized data of heart rate variability (zHRV), internal load (zSHRZ), and movement load (zML) from 11 professional male basketball players.
• Determined optimal cluster number using the elbow criterion and assessed clustering quality with the silhouette coefficient.
• Employed repeated-measures correlations to analyze associations between physiological variables.

Main Results:

• Identified three significant physiological load profiles: High load-elevated HRV, Low load-near-average HRV, and High load-reduced HRV.
• Observed substantial temporal and inter-individual variability in these profiles across the season.
• Found a strong association between internal load (SHRZ) and movement load (ML) (r=0.87), but weak correlations between HRV and SHRZ/ML.

Conclusions:

• Unsupervised clustering provides a data-driven framework for summarizing weekly physiological load states in elite athletes.
• This approach has the potential to enhance individualized load management strategies.
• Future research should explore the link between these identified profiles and performance, fatigue, and injury risk.