Scholarlink

<p>With the deep integration of information technology and sports science, data-driven physical fitness assessment methods have become an important support for the intelligent development of modern sports training. Traditional physical fitness assessment mainly relies on offline testing and experience-based judgment, making it difficult to achieve continuous monitoring and dynamic analysis of physical fitness status during exercise. To address these issues, this paper proposes an interpretable machine learning framework for sports physical fitness assessment based on the Internet of Things (IoT). This framework uses wearable sensing devices as data acquisition terminals, achieves real-time transmission and centralized management of multi-source physical fitness data through IoT communication, and completes data preprocessing and feature construction under an edge-cloud collaborative architecture. At the model level, this paper constructs a physical fitness assessment model based on a backpropagation neural network and introduces a particle swarm optimization (PSO) algorithm to globally optimize network weights and threshold parameters, forming a PSO-BP hybrid intelligent assessment model to improve model convergence speed and prediction accuracy. Simultaneously, a multi-dimensional physical fitness feature system covering physiological indicators, exercise load indicators, and energy metabolism indicators is established to characterize the complex nonlinear relationships between physical fitness states. To enhance the transparency and practicality of the model, the SHAP interpretability analysis method is further introduced to quantitatively analyze the contribution of different physical fitness characteristics to the assessment results, alleviating the “black box” problem of traditional machine learning models. Experimental results show that the proposed method outperforms traditional models in terms of accuracy, stability, and interpretability in physical fitness assessment, and can effectively reflect the dynamic changes in physical fitness during exercise.</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

An IoT-based interpretable machine learning framework for physical fitness evaluation in sports training

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月7日

查看全文
  • Lei Dang,
  • Yimeng Fan,
  • Shaobin Zhang

摘要

With the deep integration of information technology and sports science, data-driven physical fitness assessment methods have become an important support for the intelligent development of modern sports training. Traditional physical fitness assessment mainly relies on offline testing and experience-based judgment, making it difficult to achieve continuous monitoring and dynamic analysis of physical fitness status during exercise. To address these issues, this paper proposes an interpretable machine learning framework for sports physical fitness assessment based on the Internet of Things (IoT). This framework uses wearable sensing devices as data acquisition terminals, achieves real-time transmission and centralized management of multi-source physical fitness data through IoT communication, and completes data preprocessing and feature construction under an edge-cloud collaborative architecture. At the model level, this paper constructs a physical fitness assessment model based on a backpropagation neural network and introduces a particle swarm optimization (PSO) algorithm to globally optimize network weights and threshold parameters, forming a PSO-BP hybrid intelligent assessment model to improve model convergence speed and prediction accuracy. Simultaneously, a multi-dimensional physical fitness feature system covering physiological indicators, exercise load indicators, and energy metabolism indicators is established to characterize the complex nonlinear relationships between physical fitness states. To enhance the transparency and practicality of the model, the SHAP interpretability analysis method is further introduced to quantitatively analyze the contribution of different physical fitness characteristics to the assessment results, alleviating the “black box” problem of traditional machine learning models. Experimental results show that the proposed method outperforms traditional models in terms of accuracy, stability, and interpretability in physical fitness assessment, and can effectively reflect the dynamic changes in physical fitness during exercise.