Table 11 Feature selection comparison.
From: Towards advanced AI-based solutions for securing IoMT in smart health information systems
Method | Algorithm | Feature selection technique | No. selected features | Overfitting prevention | Interpretability | Features stability index* |
|---|---|---|---|---|---|---|
BiGRU/PSO | PSO | RFE with SVM | 43 | Moderate | High | 0.82 |
BiGRU/GA | GA | RFE with SVM | 45 | Moderate | High | 0.80 |
BiGRU/GWO | GWO | RFE with SVM | 42 | Strong | High | 0.85 |
BiGRU/WOA | WOA | RFE with SVM | 44 | Moderate | High | 0.81 |
BiGRU/HHO | HHO | RFE with SVM | 41 | Strong | High | 0.86 |
BiGRU/MPA | MPA | RFE with SVM | 43 | Moderate | High | 0.83 |
BiGRU/RBWK (Proposed) | RBWK | RFE with SVM | 40 | Strong | High | 0.91 |
Random Forest | N/A | Gini Importance | 50 | Moderate | Moderate | 0.75 |
LSTM | Adam | Manual Thresholding | 60 | Weak | Low | 0.62 |
CNN | Adam | Filter-Based | 70 | Weak | Low | 0.58 |
Autoencoder | Adam | Reconstruction Error | 30 | Moderate | Low | 0.68 |
CNN-LSTM Hybrid | Adam | Wrapper-Based Search | 45 | Moderate | Low | 0.70 |