Enhancing Runner Safety: A Machine Learning Framework for Injury Prediction

Avoiding injuries is crucial for athletic performance. Despite the difficulty in predicting ‎injuries, new technologies and data science applications may offer valuable information. The ‎goal was to apply machine learning to predict in juries in runners based on complete training ‎data. Over a seven-year span, 74 elite middle- and long-distance runners were used to test ‎injury prediction. Two methods of analysis were used. Initially, a time series was created that ‎represents the training load for the preceding seven days, with ten features describing the ‎training for each day. These characteristics combined subjective information on training ‎effectiveness and training effort level with data from a watch (such as duration and distance). ‎The average area under the ROC-AUC curve for the day approach were produced to be ‎‎0.9978 by a prediction system based on Random Forest Classifier machine learning model. ‎The machine learning-based method predicts a significant percentage of injuries when the ‎model is built using training data recorded on a daily basis in athlete training. All things ‎considered, the results showcase the benefits of applying machine learning to forecast injuries ‎and also customize athlete training regimens‎.