Scholarlink

AI’s Evolution in Sports Analytics: Transforming Injury Prediction Models In today’s football, injuries are one of the most serious challenges that teams have to face, since they directly impact not only the individual performance of the athlete but also the performance of the team as a whole. The recent advancements in GPS tracking systems along with deep learning methods have facilitated accurate identification of all possible injury risks even before they occur, assisting coaches, medical staff and sports scientists to adopt corrective measures. A new and improved XGBoost (Extreme Gradient Boosting) model for predicting football player injuries has been proposed for GPS tracking data, which has been used for obtaining the tracking data of Al-Talaba Sports Club players during the 2023–2024 Iraqi Premier League season. Data consists of physical and physiological parameters, specifically playing load, heart rate, collision impact and calorie expenditure, derived from PlayerTek Plus GPS devices. For validation purpose, the proposed model with the training of real matches was compared with Random Forest (RF) and Multilayer Perceptron (MLP) models, in terms of the predictor efficiency. Experimental results show that the enhanced XGBoost model achieves 95.72% recall, 91.39% accuracy, 91.74% precision, and 0.241 root mean square error (RMSE), surpassing traditional ML methods. AI-enabled models will ultimately be a dependable injury prediction tool that minimizes injury-associated downtimes, maximizes athlete fitness bandwidths, and supports training load supervision. Our findings range beyond existing literature as to when players will injure themselves, as the results can be integrated into current methods of player management, or other team strategies where player performance longevity is required thus supporting the advancement of AI-powered sports analytics. A potential next step for modelling would be the integration of deep learning architectures such as LSTMs or CNNs for injury predictions with even further accuracy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Using Enhanced XGBoost Algorithm to Predict Football Player Injuries Based on GPS Tracking Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Haitham Jawad Kadhim,
  • Saif Abdulzahra Hamzah,
  • Ahmed Farhan Ali,
  • Omar Mohammed Majeed,
  • Maab Fathi Hamzah,
  • Samer Saad Ibrahim

摘要

AI’s Evolution in Sports Analytics: Transforming Injury Prediction Models In today’s football, injuries are one of the most serious challenges that teams have to face, since they directly impact not only the individual performance of the athlete but also the performance of the team as a whole. The recent advancements in GPS tracking systems along with deep learning methods have facilitated accurate identification of all possible injury risks even before they occur, assisting coaches, medical staff and sports scientists to adopt corrective measures. A new and improved XGBoost (Extreme Gradient Boosting) model for predicting football player injuries has been proposed for GPS tracking data, which has been used for obtaining the tracking data of Al-Talaba Sports Club players during the 2023–2024 Iraqi Premier League season. Data consists of physical and physiological parameters, specifically playing load, heart rate, collision impact and calorie expenditure, derived from PlayerTek Plus GPS devices. For validation purpose, the proposed model with the training of real matches was compared with Random Forest (RF) and Multilayer Perceptron (MLP) models, in terms of the predictor efficiency. Experimental results show that the enhanced XGBoost model achieves 95.72% recall, 91.39% accuracy, 91.74% precision, and 0.241 root mean square error (RMSE), surpassing traditional ML methods. AI-enabled models will ultimately be a dependable injury prediction tool that minimizes injury-associated downtimes, maximizes athlete fitness bandwidths, and supports training load supervision. Our findings range beyond existing literature as to when players will injure themselves, as the results can be integrated into current methods of player management, or other team strategies where player performance longevity is required thus supporting the advancement of AI-powered sports analytics. A potential next step for modelling would be the integration of deep learning architectures such as LSTMs or CNNs for injury predictions with even further accuracy.