Table 2 Hyperparameter values used by the PSO-GRUGAN-IDS model for traffic abnormality detection.
Parameters | Value |
|---|---|
PSO optimizer | |
Number of particles | 10 |
Convergence threshold | 0.5 |
Inertia weight | 0.5 |
Cognitive weight | 0.8 |
Social weight | 0.8 |
Global best fitness value | Dynamic (converges during optimization) |
Iteration count | Dynamic (depends on convergence) |
GAN model | |
GR layers | GRU layer (128 units) |
Dense (512 units, relu) | |
Dense (256 units, relu) | |
Dense (128 units, relu) | |
Dense(output_dim, sigmoid) | |
GR optimizer | Adam |
learning_rate | 0.0005 |
beta_1 | 0.5 |
DR layers | GRU (128 units) |
Dense (256 units, relu) | |
Dense (128 units, relu) | |
Dense (64 units, relu) | |
Dense (1 unit, sigmoid) | |
DR optimizer | Adam |
learning_rate | 0.0002 |
beta_1 | 0.5 |
Discriminator loss function | Binary cross-entropy |
Combined model loss function | Binary cross-entropy |
GAN training epochs | 25 |
GAN batch size | 32 |
IDS model | |
Model layers | GRU(128 units) |
Dense(64 units, relu) | |
Dense(1 unit, sigmoid) | |
Optimizer | Adam |
learning_rate | 0.0005 |
beta_1 | 0.5 |
The loss function | Binary cross-entropy |
Epochs | 25 |
Batch size | 32 |
