Predicting Body mass index classification from physical fitness indicators using machine learning in South Chinese college students from 2013 to 2025

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Abstract Physical inactivity and increasing obesity among college students are major public health concerns. Although physical fitness is closely related to body mass index (BMI), the combined predictive value of multiple fitness components remains unclear. Purpose: This study investigated the association between physical fitness and BMI and evaluated the predictive performance of fitness indicators for BMI classification using machine learning (ML). Methods: A total of 83,664 first-year college students (43,734 males and 39,930 females) from Shenzhen University, China, were included (2013–2025). Fitness assessments included vital capacity, standing long jump, sit-and-reach, 50 m sprint, long-distance running (800 m in females, 1,000 m in males), and pull-ups (males) or sit-ups (females). Four supervised machine learning algorithms (Logistic Regression, Decision Tree, Random Forest, and XGBoost) were developed and compared for BMI prediction. XGBoost achieved the highest predictive performance and was subsequently used to assess feature importance, including analyses stratified by gender. Results: The model showed good predictive accuracy (75% in males, 83% in females), indicating a strong association between fitness and BMI. Pull-ups were the most important predictor in males, while vital capacity and long-distance running demonstrated high predictive importance in both genders. Conclusions: Physical fitness is closely associated with BMI, and machine learning models can effectively predict BMI using routine fitness data. These findings support the use of fitness assessments as practical tools for screening and promoting weight management among young adults.