Predictive Analysis of Future Injury Using Machine Learning

We performed predictive analysis on the athlete's physical injury by leveraging multiple machine learning algorithms with the historical features of the athlete's injury.Injury is a significant concern in professional sports.Preventing physical injury is beneficial to sustain the athlete's performance and to extend their career.Recent advances in computing technology have made significant progress in injury prevention.Unfortunately, such an application is not easy.The span of the player's physical conditions is extensive, and most importantly, the area is intensively engaged in the medical regime.Acquiring athletes' injury information is strongly restricted due to personal privacy.We hypothesized that synthetic data would be a feasible tool to elucidate the recent methodological applicability for injury prediction if the athletes' physical condition is classified with their performance.Given this assumption, we evaluated the models with various metrics and inspected which features are more important for the likelihood of future injuries.Our result shows that training intensity is the most important feature, and the average accuracy is about 0.5 regardless of the models used.Since the main goal of this study is to illustrate the capability of prediction using machine learning models, we demonstrated the whole analysis procedure, including the evaluation of results.