Table 2 Summary of architectures and hyperparameters for all five models.
Model | Architecture/Layers | Key Hyperparameters | Optimizer/LR | Activation Functions | Epochs/Batch Size | Dropout/Regularization | Evaluation Metrics |
|---|---|---|---|---|---|---|---|
MLP | Dense (32) → Dense (128) → Dense (256) → Dense (64) → Dense (32) → Dense (1) | 6 hidden layers, neurons: [32,128,256,64,32], Loss: MSE | Adam/0.02 | ReLU (hidden), Linear (output) | 200/10 | None | R², MSE, RMSE |
SVR | RBF kernel | C = 13,500, ε = 0.00001, gamma=’auto’, degree = 5, coef0 = 0.07, tol = 0.42 | – | – | – | – | R², MSE, RMSE |
RBF | Gaussian Process with kernel: 24*RBF (length_scale = 100) + WhiteKernel (noise_level = 10) | α = 15, length_scale_bounds = (0.001,1000), noise_level_bounds = (1e-5,10) | L-BFGS-B (auto) | – | – | Noise handling via WhiteKernel | R², MSE, RMSE |
CNN | Conv1D (8, 2) → Conv1D (64, 2) → Conv1D (128, 2) → Conv1D (64, 2) → Conv1D (8, 2) → Flatten → Dense (8) → Dense (1) | Kernel size = 2, filters = [8,64, 32,64,64,32,8], Loss = MSE | Adam/0.03 | ReLU (Conv & Dense), Linear (output) | 300/7 | None | R², MSE, RMSE |
GRU | GRU (32) → Dropout (0.08) → Dense (256) → Dense (128) → Dense (64) → Dense (32) → Dense (16) → Dense (32) → Dense (64) → Dense (128) → Dense (256) → Dense (1) | 1 GRU + 10 Dense layers, neurons = [32,256,128,64,32,16,32,64,128,256], Loss = MSE | Adam/default | ReLU (Dense), Linear (output) | 300/32 | Dropout = 0.08 | R², MSE, RMSE |
