Table 10 Comparative analysis of existing methods vs. proposed framework.

From: An enhancement of machine learning model performance in disease prediction with synthetic data generation

Method

Core components

Key features

Limitations

AUC (avg)

F1-score (avg)

SMOTE + TabNet

Classical Oversampling + TabNet

Fast, interpretable, lightweight

Poor minority class diversity; limited generalization

0.79

0.76

ADASYN + TabNet

Adaptive Sampling + TabNet

Adjusts for class imbalance adaptively

Generates noisy borderline samples

0.78

0.75

Deep-CTGAN + TabNet

GAN-based Generation + TabNet

Captures tabular distribution nuances

No feature-level enhancement

0.82

0.79

ResNet + TabNet

Feature Enhancement + Classifier

Deep representation learning on tabular features

Sensitive to imbalanced data distribution

0.80

0.77

Proposed (Hybrid)

Deep-CTGAN + ResNet + TabNet

End-to-end: generation, enhancement, interpretation

Higher runtime; added architectural complexity

0.85

0.81

Back to article page