Table 7 Hyperparameters Details for Proposed Hybrid Model.
Hyperparameters | Proposed Value | Description |
|---|---|---|
Learning Rate | 0.001 | Determines the step size at each iteration while moving toward a minimum during training |
Batch Size | 32 | Number of training samples used in one iteration to update model weights |
Number of Epochs | 50 | Total number of complete passes through the training dataset |
Optimizer | Adam | Optimizer used to minimize the loss function, providing efficient learning |
Dropout Rate | 0.5 | Dropout is used to prevent overfitting by randomly dropping a percentage of neurons during training |
ResNet-50 Learning Rate | 0.0001 | Learning rate for ResNet-50 to ensure stable and fine-tuned training |
EfficientNet-B0 Learning Rate | 0.0001 | Learning rate for EfficientNet-B0 to balance optimization |
DenseNet121 Learning Rate | 0.0001 | Learning rate for DenseNet121 to prevent overfitting while optimizing accuracy |
Fuzzy Logic Rule Base | 9 rules | The number of rules used for the fuzzy logic-based weighting mechanism in the ensemble |
C-GAN Latent Vector Size | 100 | Size of the latent vector used in the generator of the C-GAN |
C-GAN Batch Size | 64 | Batch size used in the C-GAN model for training the generator and discriminator |
C-GAN Epochs | 100 | Number of epochs used to train the C-GAN for generating synthetic samples |
Synthetic Data Augmentation Ratio | 0.2 | Percentage of synthetic data generated by C-GAN relative to the original dataset |
Ensemble Weighting Mechanism | Dynamic based on accuracy and confidence | The weighting mechanism of the ensemble is adjusted dynamically based on the individual models’ performance |
