Supervised Machine Learning Algorithms for Fitness-Based Cardiometabolic Risk Classification in Adolescents.

Area of Science:

• Public Health and Preventive Medicine
• Computational Biology and Bioinformatics
• Adolescent Health and Sports Science

Background:

• Adolescent cardiometabolic risk is a significant public health issue.
• Physical fitness is a key modifiable factor influencing cardiometabolic health.
• Traditional statistical methods struggle with complex relationships in fitness and anthropometric data.

Purpose of the Study:

• To develop and evaluate supervised machine learning algorithms for classifying adolescent cardiometabolic risk.
• To utilize standardized physical fitness assessments as input for risk prediction models.
• To compare the performance of various machine learning models in identifying at-risk adolescents.

Main Methods:

• Cross-sectional analysis of a representative sample of school-aged adolescents.
• Inclusion of field-based physical fitness tests: cardiorespiratory fitness (VO2max), muscular strength (push-ups), and explosive power (horizontal jump).
• Application and comparison of supervised machine learning models (e.g., artificial neural networks, ensemble methods) using accuracy, F1 score, recall, and AUC-ROC metrics.

Main Results:

• The gradient boosting classifier demonstrated superior performance among tested models.
• Achieved 77.0% accuracy, 67.3% F1 score, and the highest AUC-ROC (0.601), indicating effective risk classification.
• Horizontal jumps and push-up performance were identified as the most significant predictive variables for cardiometabolic risk.

Conclusions:

• Gradient boosting is a highly effective model for predicting adolescent cardiometabolic risk from physical fitness data.
• This machine learning approach provides a practical, data-driven tool for early risk detection in adolescents.
• The findings support the potential for scalable screening programs in educational and clinical settings.