A sequential deep learning model for predicting people with obesity in adults aged 18–64 using physical fitness variables
Authors' conclusion
This explainable sequential deep learning model accurately predicts people with obesity using physical fitness variables and highlights the critical role of cardiorespiratory fitness in obesity risk assessment and management.
Study context
- Population
- Adults with nationally representative physical fitness data, collected between 2010 and 2023
- Participants
- 204,334 adults
- Outcome
- Prediction of obesity
Methodological detailssequential deep learning model, stratified 5-fold cross-validation, explainable (SHAP and LIME), nationally representative data
- Sequential deep learning model
- Stratified 5-fold cross-validation
- Explainable (SHAP and LIME)
- Nationally representative data
Performance
As reported in the abstract
- Overall accuracy across five stratified folds (p < 0.001 vs. random)
- 0.87-0.88
- Fold 4 accuracy
- 0.873
- Fold 4 precision
- 0.866
- Fold 4 recall
- 0.855
- Fold 4 F1
- 0.876
- Fold 4 ROC-AUC
- 0.95
- Fold 4 TN
- 1,462
- Fold 4 FN
- 184
- Fold 4 FP
- 249
- Fold 4 TP
- 1,554
Further findings
Error metrics (fold 4 model)
- MAE
- 0.122
- MSE
- 0.041
- R²
- 0.833
- Average prediction error, first 50 samples
- 0.171
SHAP feature importance (top predictors)
- 20-m shuttle run estimated VO₂max
- 0.339
- Gender
- 0.2481
- Relative grip strength
- 0.135
LIME interpretation
- LIME intercept
- 0.511
- LIME predicted
- 0.668
- LIME R²
- 0.995
- Low estimated VO₂max (< 28.71 ml/kg/min) and low relative grip strength (< 38.17%) substantially increased probability of obesity classification, particularly among females
Original publication
A sequential deep learning model for predicting people with obesity in adults aged 18–64 using physical fitness variables. Int J Obes (Lond). Published 20 March 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →