Технологии машинного обучения при определении факторов, связанных с долей жира в организме Вьетнамских подростков

Abstract

The aim of this study was to investigate the factors influencing Body Fat Percentage (BFP) among Vietnamese adolescents aged 11 to 15 employing machine learning techniques for predictive analysis. A total of 1,208 adolescents, comprising 598 boys and 610 girls, drawn from nine junior high schools in Vietnam's capital, were enrolled in the study. Body composition measurements were conducted using the HBF 375 (Omron) device by Bioelectrical Impedance Analysis method. The study questionnaire, initially validated by The National Institute of Nutrition, encompassed inquiries related to dietary behaviors, meal frequencies, physical activities, sedentary habits, and nutritional knowledge. A machine learning methodology employing a decision tree algorithm was employed to discern the primary determinants most significantly correlated with BFP. This study successfully identified six distinct predictor groups associated with BFP among adolescents, leveraging the decision tree model, with Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) values of 4.80 and 3.80, respectively. Among these predictors, frequency of fruit consumption, snacking habits, mode of transportation to school, and screen time (computer and/or cell phone usage) emerged as the most influential factors linked to BFP in adolescents. The combination of these factors and interactions with gender and pubertal status can BFP in Vietnamese adolescents. This research sheds light on the complex and diverse factors impacting BFP in Vietnamese adolescents. This study's results underscore the practical importance of promoting healthy eating and exercise habits among adolescents, offering valuable insights for parents and schools to enhance their childcare strategies.