Classification of vertical jump performance categories in futsal using machine learning algorithms.

Area of Science:

• Biomechanics
• Sports Science
• Machine Learning

Background:

• Futsal demands explosive vertical jumps, generating complex force-time data.
• Traditional analysis struggles with multidimensional biomechanical data from dual force plates.

Purpose of the Study:

• Develop a supervised classification framework for futsal athletes' vertical jump performance.
• Utilize independent force-time metrics from dual force plates for classification.

Main Methods:

• Fifty-one male athletes performed countermovement jumps, squat jumps, and drop jumps.
• Principal Component Analysis (PCA) derived four performance categories from six biomechanical variables.
• Four machine learning algorithms were compared using athlete-level stratified cross-validation.

Main Results:

• Logistic Regression achieved high accuracy (F1-Score=0.830, AUC-ROC=0.977) on the independent test set.
• Cross-validation provided conservative estimates (F1=0.770 ± 0.069, AUC=0.941 ± 0.039).
• Countermovement jump peak power and drop jump variables were most influential predictors.

Conclusions:

• Machine learning, specifically Logistic Regression, accurately classifies futsal athletes' vertical jump performance.
• The PCA and supervised classification framework offers a reproducible method for talent identification and training.
• Rigorous validation demonstrates the framework's potential for longitudinal monitoring and individualized training programs.