Developing new VO₂max prediction models from maximal, submaximal and questionnaire variables using support vector...

Area of Science:

• Exercise Physiology
• Machine Learning in Sports Science

Background:

• Maximal oxygen uptake (VO2max) is a crucial indicator of cardiorespiratory fitness and overall health.
• Accurate prediction of VO2max is vital for personalized training and health assessments.

Purpose of the Study:

• To develop novel VO2max prediction models by integrating maximal, submaximal, and questionnaire variables.
• To identify the most significant predictors of VO2max using Support Vector Machines (SVM) and Relief-F feature selection.

Main Methods:

• Utilized Support Vector Machines (SVM) combined with the Relief-F feature selector for VO2max prediction.
• Developed hybrid models using double and triple combinations of maximal, submaximal, and questionnaire variables.
• Employed 10-fold cross-validation to evaluate model performance using R and RMSE metrics.

Main Results:

• The best prediction model, combining triple relevant variables, achieved R=0.94 and RMSE=2.92mLkg-1min-1.
• Key predictors identified include gender, age, maximal heart rate (MX-HR), submaximal ending speed (SM-ES), and Perceived Functional Ability (Q-PFA).
• SVM models demonstrated superior accuracy compared to Multilayer Perceptron (MLP) and Tree Boost (TB) regression methods.

Conclusions:

• The novel SVM-based approach integrating diverse data sources significantly enhances VO2max prediction accuracy.
• The identified predictors offer valuable insights for understanding and assessing cardiorespiratory fitness.
• This study establishes a new benchmark for VO2max prediction models in sports science and health.