Energy availability is defined as the amount of available energy to support all other bodily systems after subtracting the energy expended during training from the total energy intake, considering the energy consumed during exercise. A prolonged state of low energy availability can lead to the development of Relative Energy Deficiency in Sport (REDs) syndrome and other serious health issues, as well as impact athletic performance, especially relevant for dancers exposed to intense physical demands. The Dance-Specific Energy Availability Questionnaire (DEAQ) has been developed to identify dancers at increased risk of developing REDs. At the Faculty of Sport, the reliability and validity of this questionnaire are being actively researched. For this purpose, a dataset was compiled, which includes data on numerous dancers, including morphological and blood measurements, as well as responses to the questionnaire. The dataset was cleansed and rebalanced using the ADASYN (Adaptive Synthetic Sampling Approach for Imbalanced Learning) method. The most suitable attributes for learning were selected using the ReliefF method. We then built predictive models using different Machine learning methods. The most accurate model, built using the gradient boosting method, achieved a classification accuracy of 0.9071. The results suggest that machine learning models could be a useful tool for screening low energy availability in female dancers of different dance styles.
|
