V. Uluslararası Uygulamalı İstatistik Kongresi (UYIK - 2024) İstanbul / Türkiye, 21-23 Mayıs 2024, İstanbul, Türkiye, 21 - 23 Mayıs 2024, ss.270-280
The physical health of athletes participating in competitive sports events significantly affects their success in
competitions. On the other hand, with the industrialization of sports, the increase in investments in this field brings
the expectation of success. Thus, a significant part of the investments is spent on protecting the physical health of
athletes, who are the most critical actors in this field. In this context, accurate prediction of injuries that may occur
by considering the various characteristics of athletes is essential for both the players' sustainable career planning
and the teams' goals. However, as well as predicting the injury status of athletes with high accuracy, it is also
essential that the model used is explainable and that the features that most affect the prediction results are revealed.
In this way, the action to be taken by the athlete, who is predicted to have a possibility of injury, and the decisions
that the club managers will make about the athlete can be determined. In this study, the likelihood of the player
experiencing an injury will be predicted using various machine learning models, considering factors such as
athletes' physical characteristics, injury history, training intensity, and recovery time. The models used will be
compared, considering their accuracy of prediction as well as their explainability. Thanks to the comparative
analyses presented as a result of the study, guiding results will be provided for decision-makers to choose the most
suitable one among models with different prediction accuracy and explainability levels.