A Novel Superaug Adaptive Data Augmentation Strategy for Metabolic Syndrome Prediction: The First Egyptian National Cohort Study With Multi-National Comparison

Nihal Elsonny, Biomedical Engineering Program, Faculty of Engineering, Mansoura University, Mansoura, EgyptFollow
Ashraf Elsharkawy, Pediatric Department, Head of Endocrine & Diabetes Unit, Faculty of Medicine, Mansoura University, Mansoura, Egypt
Abeer Tawakol, Electronics and Communications Engineering Department, Faculty of Engineering, Mansoura University, Mansoura, Egypt
Amira Y. Haikal, Computers and Control Systems Engineering Department, Faculty of Engineering, Mansoura University, Mansoura, Egypt

Subject Area

Biomedical Engineering

Article Type

Original Study

Abstract

Metabolic Syndrome (MetS) is a multifactorial disorder associated with an increased risk of cardiovascular disease, type 2 diabetes, and obesity-related complications. Early and accurate prediction of MetS remains challenging due to heterogeneous data sources, class imbalance, limited sample sizes in regional studies, and poor generalization across populations.

This study introduces a comprehensive machine learning framework for predicting MetS and its prognostic factors using two heterogeneous datasets: a newly collected Egyptian clinical cohort and the U.S.-based (NHANES) [CDC, 2023] dataset. For each dataset, multiple experimental pipelines are developed, integrating preprocessing, feature selection, data augmentation, hyperparameter tuning, and ensemble modeling.

A key contribution of this work is SUPERAUG (Supervised Ensemble and Robust Augmentation), an adaptive and hybrid augmentation strategy that combines multi-noise injection with SMOTE [Chawla et al., 2002] to enhance minority class representation and improve data diversity. The proposed approach is evaluated in both standard and adaptive configurations, where augmentation strength is dynamically adjusted based on model performance.

To ensure robust evaluation, repeated cross-validation and statistical analysis are employed, particularly for the small-sized Egyptian dataset. Experimental results demonstrate that SUPERAUG improves model generalization and class balance across both datasets. The best-performing pipelines achieved up to 93.75% accuracy on the Egyptian dataset and 91.79% on the NHANES dataset, while more conservative configurations yield stable and reliable performance.

Across both populations, waist circumference, fasting glucose, triglycerides, and body mass index (BMI) consistently emerged as the most predictive features. These findings highlight the adaptability and effectiveness of the proposed framework and support its potential for population-specific clinical decision support systems.

Keywords

Machine Learning; Metabolic Syndrome; Data Augmentation; Ensemble Learning; SUPERAUG; MetS; Predictive Analytics

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.

Recommended Citation

Elsonny, Nihal; Elsharkawy, Ashraf; Tawakol, Abeer; and Haikal, Amira Y. (2026) "A Novel Superaug Adaptive Data Augmentation Strategy for Metabolic Syndrome Prediction: The First Egyptian National Cohort Study With Multi-National Comparison," Mansoura Engineering Journal: Vol. 51 : Iss. 5 , Article 5.
Available at: https://doi.org/10.58491/2735-4202.3460

Download