Predictive models for injury prevention in football: a machine learning approach using performance and injury data

Injuries in professional football represent a critical issue that significantly affects both team performance and the financial sustainability of clubs. This thesis investigates whether, based on data concerning players’ on-field performance and historical injury records, it is possible to develop an effective predictive model to anticipate the risk of injury in the weeks following a match. The work begins with a structured review of the existing scientific literature on injury prediction in sports, with particular focus on machine learning applications in football. Subsequently, a dataset was constructed by integrating information from publicly available sources (Fbref and Transfermarkt), encompassing detailed match statistics and injury timelines. Machine learning algorithms were then applied to this dataset to train and evaluate predictive models. The aim was not only to assess the accuracy and reliability of these models, but also to identify the most influential features in predicting injury risk, thereby providing actionable insights for clubs and medical staff engaged in injury prevention strategies.