Scholarlink

<p>Machine learning and computer vision allow the development of sophisticated models for evaluating an athlete's readiness and fatigue. In this paper, we studied the effects of stressors faced by athletes to comprehensively evaluate their readiness and fatigue while maximizing their game performance and minimizing the risk of injury. An athlete's readiness and fatigue were quantified using a modified reactive strength index (<i>RSImod</i>), computed using countermovement vertical jumps. Our study was conducted over 26&#xa0;weeks with 17 collegiate women's basketball athletes. The proposed model first learns the relationship between&#xa0;<i>RSImod</i>&#xa0;and the athletes' physical, physiological, and cognitive features. Then, it augments its learning by considering the&#xa0;smallest worthwhile change (SWC) of the five most significant features that correlated well with <i>RSImod</i>&#xa0;to account for intra-athlete variability. Finally, we used our proposed hierarchical approach employing decision tree classifiers and regressors (ensemble–boosting) to predict an athlete's <i>RSImod</i> score for the following week. Our experiments demonstrated that SWC augmentation improved <i>RSImod</i> level prediction accuracy from 92.83% (original dataset) to 95.28%. The proposed hierarchical approach performs better (MSE 0.011, R<sup>2</sup>&#xa0;0.963) than state-of-the-art prediction algorithms (multilinear and random forest regressor), generates interpretable outcomes, and helps coaches develop effective training schedules and game strategies. When tested without SWC augmentation, the hierarchical model achieved an MSE of 0.028 and an adjusted R<sup>2</sup> of 0.906. SWC augmentation reduced the MSE by 60.71% (from 0.028 to 0.011). It increased the adjusted R<sup>2</sup> by 6.29% (from 0.906 to 0.963), further highlighting the combined efficacy of SWC augmentation and the hierarchical approach. By integrating various physical, physiological, and cognitive features, the proposed model helps coaches optimize athlete performance and mitigate injury risks effectively.</p>

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

Self-explaining Hierarchical Model for Fatigue Monitoring and Prediction in Basketball Self-explaining Hierarchical Model for Fatigue Monitoring and Prediction in Basketball

  • 期刊论文

  • 发表日期：

    2025年2月12日

查看全文
  • Srishti Sharma,
  • Srikrishnan Divakaran,
  • Tolga Kaya,
  • Mehul S. Raval

摘要

Machine learning and computer vision allow the development of sophisticated models for evaluating an athlete's readiness and fatigue. In this paper, we studied the effects of stressors faced by athletes to comprehensively evaluate their readiness and fatigue while maximizing their game performance and minimizing the risk of injury. An athlete's readiness and fatigue were quantified using a modified reactive strength index (RSImod), computed using countermovement vertical jumps. Our study was conducted over 26 weeks with 17 collegiate women's basketball athletes. The proposed model first learns the relationship between RSImod and the athletes' physical, physiological, and cognitive features. Then, it augments its learning by considering the smallest worthwhile change (SWC) of the five most significant features that correlated well with RSImod to account for intra-athlete variability. Finally, we used our proposed hierarchical approach employing decision tree classifiers and regressors (ensemble–boosting) to predict an athlete's RSImod score for the following week. Our experiments demonstrated that SWC augmentation improved RSImod level prediction accuracy from 92.83% (original dataset) to 95.28%. The proposed hierarchical approach performs better (MSE 0.011, R2 0.963) than state-of-the-art prediction algorithms (multilinear and random forest regressor), generates interpretable outcomes, and helps coaches develop effective training schedules and game strategies. When tested without SWC augmentation, the hierarchical model achieved an MSE of 0.028 and an adjusted R2 of 0.906. SWC augmentation reduced the MSE by 60.71% (from 0.028 to 0.011). It increased the adjusted R2 by 6.29% (from 0.906 to 0.963), further highlighting the combined efficacy of SWC augmentation and the hierarchical approach. By integrating various physical, physiological, and cognitive features, the proposed model helps coaches optimize athlete performance and mitigate injury risks effectively.