To develop a machine learning-ready dataset for predicting running-related injuries (RRI) using a variety of risk factors from multiple disciplines.
Key Findings:
Moderate improvement in prediction performance (AUC = 0.784 ± 0.014) compared to previous models, indicating a better predictive capability.
Random forest algorithm achieved the highest performance (AUC = 0.781 ± 0.016, 0.784 ± 0.014).
Logistic regression showed significant improvement when trained with a broader range of risk factors.
Interpretation:
The study provides a reproducible framework for future machine learning research in sports injury prediction, emphasizing the importance of integrating diverse risk factors to enhance predictive accuracy.
Limitations:
The dataset is limited to competitive endurance runners, which may affect generalizability to other populations.
Some metrics need to remain normalized to ensure participant anonymity, potentially limiting data richness.
Conclusion:
This research contributes valuable insights and a dataset for advancing machine learning applications in predicting running-related injuries.
