MACHINE LEARNING APPROACHES TO IDENTIFY CLUSTERS OF MOTOR COMPETENCE IN CHILDREN AGED 6–7 YEARS

Abstract

Motor competence (MC) is a key determinant of health and physical activity in childhood. Machine-learning techniques can clarify how individual, behavioural, and contextual factors shape MC profiles. This study aimed to identify Clusters (C) of MC in children aged 6–7 years and to examine their associations with physical fitness, body composition, sedentary behaviour, screen time, sleep, and school/housing context. A sample of 373 children (50%girls) was assessed. MC was classified using an unsupervised machine-learning algorithm (Two-Step Cluster analysis of Motor Competence Assessment, MCA, total percentile), which yielded three groups based on the 348 valid cases: C1-high competence (MCA range=58-91, Mean=68, n=135), C2- medium competence (MCA range=33-57, Mean=46, n=159), and C3-low competence (MCA range=6-32, Mean=21, n=54). Predictors of cluster membership were then tested using a supervised machine-learning approach (multinomial logistic regression) in a subsample of 118 children with complete data, including age, sex, BMI, waist-to-height ratio, 20m shuttle run test (SR_20m), moderate-to-vigorous physical activity (MVPA), sedentary time, screen time, sleep parameters, and rural/urban school and housing context. The supervised regression model significantly predicted cluster membership (χ²(26)=41.351, p=.029; Nagelkerke R²=.347). The SR_20m emerged as the strongest predictor (p=.007), with each additional repetition increased the odds of belonging to the C1 compared to the C3 by 18% (OR=1.18; 95%CI:1.05-1.33). Sedentary behaviour also differentiated clusters, with higher sedentary time was associated with a greater likelihood of low competence (p=.035; OR=1.30; 95%CI:1.02-1.66). No significant effects were observed for sex, BMI, MVPA, screen time, sleep, or contextual factors. The integration of unsupervised and supervised machine-learning methods identified distinct MC clusters, with the cardiorespiratory fitness and sedentary behaviour emerging as the main predictors. These findings highlight the importance of promoting physical fitness and reducing sedentary time to enhance MC in children.