Scholarlink

Elite Gaelic football demands rapid transitions between high-intensity efforts and recovery, yet accurately forecasting these changes in intensity remains a significant challenge for coaches and sport scientist. This study introduces a novel framework for forecasting intensity in elite football players by combining wearable sensor data with machine learning techniques. The Metabolic Work Ratio (MWR) is introduced as a new metric derived from metabolic power calculations, which captures the balance between high-intensity efforts and recovery within fixed time windows. Features derived from 3-min windows of wearable sensor data were extracted and used to train machine learning models to forecast transitions in MWR. Model interpretability was enhanced through SHAP analysis, identifying the magnitude and influence of the most impactful features. Our results demonstrate the feasibility of forecasting transitions between high-intensity and recovery phases, offering a proof-of-concept for data-driven methodologies that could support coaches and sport scientists in monitoring player workload and informing substitution decisions.

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

Forecasting High-Intensity Activity in Elite Gaelic Football: A Machine Learning Approach Using Wearable Sensor Data and Metabolic Work Ratio

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Valerio Antonini,
  • Mark Roantree,
  • Michael Scriney

摘要

Elite Gaelic football demands rapid transitions between high-intensity efforts and recovery, yet accurately forecasting these changes in intensity remains a significant challenge for coaches and sport scientist. This study introduces a novel framework for forecasting intensity in elite football players by combining wearable sensor data with machine learning techniques. The Metabolic Work Ratio (MWR) is introduced as a new metric derived from metabolic power calculations, which captures the balance between high-intensity efforts and recovery within fixed time windows. Features derived from 3-min windows of wearable sensor data were extracted and used to train machine learning models to forecast transitions in MWR. Model interpretability was enhanced through SHAP analysis, identifying the magnitude and influence of the most impactful features. Our results demonstrate the feasibility of forecasting transitions between high-intensity and recovery phases, offering a proof-of-concept for data-driven methodologies that could support coaches and sport scientists in monitoring player workload and informing substitution decisions.