Date of Award
Open Access Thesis
It is of utmost importance to sports organizations that they keep their players as healthy as possible and contributing to the success of the team. Advancements in technology and investments by sports clubs have allowed researchers to better understand the role of load management in high-level athletes to mitigate injury risk. Through GPS tracking data provided by a collaborating Division I American college football team, we seek to predict lower body soft tissue injuries in future training sessions and reduce the number of potentially avoidable injuries within the organization. The difficulty of analyzing the injury data set is that the frequency of injury is low compared to the noninjury cases, resulting in highly imbalanced classes. We address this imbalance issue through under or oversampling the underrepresented class by bagging, and we compare various classification procedures such as random forests, penalized logistic regression and support vector machines. Our empirical results show that the random forest classifier with undersampling outperforms other methods with respect to precision-recall (PR) and ROC curves.
Tice, N.(2022). Predicting Lower Body Soft Tissue Injuries in American Football with GPS Data. (Master's thesis). Retrieved from https://scholarcommons.sc.edu/etd/6847