•  
  •  
 

Document Type

Abstract

Publication Date

1-14-2026

Abstract

University students represent the core driving force behind the modernization of a nation. Their physical fitness is not only crucial for individual development but also serves as a key indicator of national population quality and sustainable development. Since the implementation of policies such as the National Student Physical Fitness Standards (2014 Edition), the overall physical fitness of Chinese university students has significantly improved. However, rising obesity rates and declining physical fitness levels among students remain pressing concerns. The rapid advancement of artificial intelligence (AI) technologies offers new pathways to address challenges such as inefficient health data management and the lack of scientific health planning for university students. This study employs Python to construct a neural network model utilizing student physical fitness data collected from Hubei University between 2014 and 2024. A convolutional neural network (CNN) model is developed, adhering strictly to the CRISP-DM methodology throughout the modeling and data analysis process. Furthermore, SHAP values are introduced to analyze the neural network's output, enabling an in-depth examination of the characteristics contributing to low PFI scores among different gender-body type groups. 1) The proportion of individuals with low PFI scores within different BMI groups varied by gender. Except for normal-weight males (69.51%), all other groups exhibited proportions exceeding 70%. 2) The relationship between BMI and PFI among university students is nonlinear, with PFI peaking when BMI approaches the normal range. 3) The contributing factors to low PFI scores differ across gender-BMI subgroups, emphasizing the need for tailored physical fitness improvement strategies based on individual body types. Analyzing health data from different gender-body type groups using SHAP values is a feasible approach. The explainable deep learning model based on SHAP overcomes the limitations of traditional regression analysis and accurately identifies the specific weaknesses of low-PFI individuals. For instance, overweight males aiming to enhance their physical fitness should prioritize body composition improvement and upper and lower limb strength development. It is recommended that universities implement targeted intervention programs based on BMI and gender stratification.

DOI

https://doi.org/10.18122/ijpah.5.1.178.boisestate

Download

Share

COinS