Athlete-customized injury prediction using training load statistical records and machine learning‏

Abstract

The management of athletic performance is of immense importance in the sports industry. Performance management is concerned with maximizing athletes' performance and minimizing the risk of player injuries. Several factors are contributing to those objectives, including player health status, emotional conditions (e.g., stress and anxiety), athletic load and physical demands (e.g., jumping and landing tasks), etc. Generally, the prediction of injury is a key component for injury prevention as the successful identification of injury predictors forms the basis for effective preventive measures. This study seeks to develop and validate a hierarchical machine learning predictive system that possess the ability to make an early and accurate detection of a player's injury using athletic load data. This early personalize injury detection can help in injury avoidance by identifying required physical and workloads. Our framework was tested on athletic data for 21 soccer players that was collected and/or measured from different sources including internal load data (such as heart rate), external load data (such as the duration of workout and number of jumps), as well as questionnaire data. All these data are integrated into the proposed system to increase injury prediction accuracy. The proposed algorithm identifies players at risk of injury so that early interventions can be made.