Prediction of maximum oxygen uptake from maximal and non-exercise variables using machine learning methods

The purpose of this thesis is to develop accurate VO2max prediction models using maximal and N-Ex variables and show that the accuracy of maximal prediction models can be improved by adding N-Ex questionnaire predictor variables. Machine learning methods such as SVM, ANN, MLR and GRNN are used for developing VO2max prediction models.Two different datasets are used for VO2max prediction.Using 10-fold cross validation, the performance of the models are evaluated by calculating R?s and SEE?s.It is shown that adding questionnaire to ANN-based prediction models yields 4.25% to 10.81% increments in R?s and 27.53% to 35.59% decrements in SEE?s for the first dataset and 1.08% to 6.97% increments in R?s and 0.53% to 17.97% decrements in SEE?s for the second dataset.Also it is concluded that ANN-based prediction models are more accurate than GRNN-based, SVM-based and MLR-based prediction models.