BMJ public health · 2026
Generative AI-enhanced ensemble model for predicting adolescent obesity in India: a school-based study in two southern districts of Karnataka.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Rakshitha Jayashankar, Sunil Kumar Doddaiah, Santhosh Kumar M, Arun Vanishri, Krishnamurthy Kamidi Venkatesh
Аннотация
BACKGROUND: Adolescent obesity is a growing concern in India, shaped by sociodemographic, behavioural, familial, sleep and mental health factors. Validated AI-based prediction models for adolescents are scarce. Generative AI (GenAI) offers opportunities to improve accuracy through synthetic data augmentation.
METHODS: A case-control, school-based study included 1292 adolescents aged 10-16 years from Mysuru and Chamarajanagar, selected via stratified random sampling. Data on sociodemographic, lifestyle, sleep, mental health, familial and anthropometric measures were collected using validated tools. The dataset was split into training (80%), validation (10%) and testing (10%). Conditional tabular generative adversarial network (CTGAN) generated 1000 synthetic samples and a variational autoencoder (VAE) produced 500, expanding training to 2532 records. Fidelity was assessed using Kolmogorov-Smirnov tests, principal component analysis (PCA), Wasserstein distance and privacy risk checks. A stacking ensemble combining XGBoost, TabNet and TabTransformer with logistic regression stacking was trained using 10-fold cross-validation.
RESULTS: The final model achieved an accuracy of 0.977, sensitivity 0.97, specificity 0.985, F1 score 0.97 and AUC 0.994. Calibration was strong (Brier score 0.042; Hosmer-Lemeshow p>0.05). SHAP and TabNet analyses highlighted screen time, sleep duration, parental obesity, physical activity and diet as key predictors.
CONCLUSION: The GenAI-enhanced stacking ensemble model accurately predicts adolescent obesity, with strong calibration and interpretability. It identifies behavioural and familial determinants, supporting school and community deployment for early detection and prevention.
METHODS: A case-control, school-based study included 1292 adolescents aged 10-16 years from Mysuru and Chamarajanagar, selected via stratified random sampling. Data on sociodemographic, lifestyle, sleep, mental health, familial and anthropometric measures were collected using validated tools. The dataset was split into training (80%), validation (10%) and testing (10%). Conditional tabular generative adversarial network (CTGAN) generated 1000 synthetic samples and a variational autoencoder (VAE) produced 500, expanding training to 2532 records. Fidelity was assessed using Kolmogorov-Smirnov tests, principal component analysis (PCA), Wasserstein distance and privacy risk checks. A stacking ensemble combining XGBoost, TabNet and TabTransformer with logistic regression stacking was trained using 10-fold cross-validation.
RESULTS: The final model achieved an accuracy of 0.977, sensitivity 0.97, specificity 0.985, F1 score 0.97 and AUC 0.994. Calibration was strong (Brier score 0.042; Hosmer-Lemeshow p>0.05). SHAP and TabNet analyses highlighted screen time, sleep duration, parental obesity, physical activity and diet as key predictors.
CONCLUSION: The GenAI-enhanced stacking ensemble model accurately predicts adolescent obesity, with strong calibration and interpretability. It identifies behavioural and familial determinants, supporting school and community deployment for early detection and prevention.