The role of morphometric characteristics in predicting 20-meter sprint performance through machine learning

Area of Science:

• Sports Science
• Biomechanical Analysis
• Machine Learning in Pediatrics

Background:

• Children's athletic performance is influenced by various morphometric factors.
• Understanding these factors is crucial for talent identification and training programs.
• Machine learning offers novel approaches to analyze complex relationships in pediatric sports science.

Purpose of the Study:

• To investigate morphometric predictors of 20-meter sprint performance in primary school children.
• To evaluate the effectiveness of different machine learning (ML) techniques and feature selection methods.
• To identify the most influential anthropometric features for predicting sprint ability in children aged 6-11 years.

Main Methods:

• Collected demographic, anthropometric (skinfold, diameter, circumference, lengths), and 20-m sprint data from 282 children (6-11 years).
• Employed three ML experiments: full feature space, correlation-selected features, and Principal Component Analysis (PCA) for dimensionality reduction.
• Utilized regression models to predict sprint performance based on selected morphometric features.

Main Results:

• Correlation analysis identified Age, Height, waist circumference, hip circumference, leg length, thigh length, and foot length as significant predictors.
• The model using correlation-based selected features achieved a minimum Mean Squared Error (MSE) of 0.012.
• This indicates a strong linear association between these selected morphometric features and sprint performance.

Conclusions:

• Morphometric features, particularly age, height, and specific body measurements, are strong linear predictors of 20-m sprint performance in children.
• Correlation analysis is an effective method for selecting relevant features in ML models for pediatric sports science.
• These findings can inform targeted training and performance analysis in young athletes.