Leveraging Lifestyle Behaviors: Identifying Risk Factors and Constructing a Predictive Model for Adolescent students Obesity in China

Mei Jiang¹Zhou He^2*

• ¹Chengdu University of Technology, Chengdu, China
• ²Southwest Jiaotong University, Chengdu, China

Adolescent obesity has emerged as a critical global public health challenge. We aimed to employ machine learning algorithms to identify significant contributing factors and develop a predictive model for adolescent obesity. The study used an anonymised adolescent dataset (n=2,338) to construct a predictive model incorporating variables such as family factors, lifestyle behaviours, and physical fitness scores. The model utilized LASSO (Least Absolute Shrinkage and Selection Operator) regression with k-fold cross-validation for variable selection, followed by logistic regression incorporating the selected variables for parameter estimation. The optimal classification threshold was determined using the Youden Index for prediction. The results indicated that the final predictors incorporated into the model included gender, mother's educational level, parental BMI, weight at age 12, parenting style, frequency of sweets consumption per week, meal duration, sleep duration, and physical fitness score. The constructed predictive model demonstrated robust performance, achieving an area under the curve (AUC) of 0.91, with an accuracy of 0.86 and a sensitivity of 0.84. Furthermore, subgroup analysis revealed that the model maintained consistent performance across genders, with a modestly superior predictive efficacy observed in the male subgroup (AUC=0.912) compared to the female subgroup (AUC = 0.898). The proposed interpretable framework integrates high predictive accuracy and sensitivity with utility, providing a valuable tool for the timely identification and intervention of high-risk adolescents. These findings highlight the potential of data-driven approaches in adolescent obesity intervention.