Professor Karl Schmitt
Arts and Sciences
Valparaiso University Women’s Soccer Organization collects heart rate data using Firstbeat Trackers to give players biofeedback on their performance, health, and recovery. However, this information only provides an opaque view into what factors are affecting their recovery. Various studies concur that the intensity of exercise has the greatest influence on recovery; but there is little published research that has analyzed the effects of environmental conditions on athletic activities. Current research shows that temperature influences heart rate variability, a commonly used metric of recovery, in non-physically demanding activities; however, there are no substantial studies that have developed predictive models that pair this with athletic exercise while controlling for intensity. Using the biometrics collected by the activity trackers, multiple machine learning algorithms and statistical methods were implemented to answer two main questions: 1) How does environmental temperature affect players’ recovery? 2) Which machine learning models and measurements of recovery provide a predictive capability that coaches can implement for better decision making? This research compares linear and non-linear modeling methods on team and individual player data and their accuracy and effectiveness of predicting athlete recovery. Models range from transparent linear regressions to black-box convolutional neural networks to find a balance between predictive capability and transparency of interpretation. Initial results show that modeling individual player data using random forest regression gives an accurate view of how factors such as temperature and intensity influence soccer athlete recovery.
Young, Mark, "Optimizing Recovery Conditions in Collegiate Female Soccer Athletes Using Machine Learning" (2020). Symposium on Undergraduate Research and Creative Expression (SOURCE). 906.