Data-driven approach to assessing performance classification of physical fitness in preadolescent children

Background: Physical fitness is a critical indicator of health and development among children; however, variation in performance across different physical metrics demands detailed classification to better understand and target interventions. 
Objectives: To classify 12-year-old children into distinct performance groups based on physical fitness metrics and to evaluate the effectiveness of these metrics in distinguishing the groups.
Method: A total of 462 children with a mean age of 12.5 ± 0.5 years was assessed across 8 fitness variables: body mass index (BMI), 1-minute curl-up, sit and reach test, handgrip test, 30-metre run, agility t-test, and standing broad jump test. K-means clustering was applied to group children into high performance (HPG), moderate performance (MPG), and low performance (LPG) groups. Discriminant analysis (DA) was used to identify significant variables to differentiate and validate classification accuracy through cross-validation in standard, forward, and backward stepwise modes.
Results: K-means clustering revealed distinct fitness profiles, with HPG characterized by high performance across most metrics, MPG demonstrating moderate performance with strengths in agility, and LPG showing weaker overall performance but excelling in speed. DA identified the 30-metre sprint and t-test as the most effective discriminators among groups, with all variables showing significant contributions (p<0.0001). Cross-validation demonstrated robust classification accuracy of 95.5% across all modes, with HPG, MPG, and LPG achieving individual accuracies of 93.9%, 96.7%, and 95.3%, respectively.
Conclusions: This study highlights utility of k-means clustering and DA in classifying fitness performance among children. Key variables like 30-metre sprint and t-test effectively distinguish performance levels providing insights for targeted fitness interventions and policy development.