Table 4 Hyperparameter tuning results and model performance from the 3 × 3 nested CV method
Model | The best parameters of the first hidden layer | The best parameters of the second hidden layer | The best parameters of the third hidden layer | Correlation coefficient on the test dataset | Mean squared error (MSE) on the test dataset |
|---|---|---|---|---|---|
1 | tanh; 512 nodes | ReLU; 256 nodes | ReLU; 512 nodes | 0.63 | 0.7 |
2 | tanh; 512 nodes | ReLU; 256 nodes | ReLU; 512 nodes | 0.61 | 0.7 |
3 | tanh; 512 nodes | ReLU; 512 nodes | ReLU; 128 nodes | 0.62 | 0.62 |