Table 3 Summary of the stepwise ANN parameters
Parameter | Setting |
|---|---|
ANN architecture | I-2-1. For each run, the increment of 1 node in the input layer, I |
Search method | Stepwise |
ANN learning algorithm | Backpropagation |
ANN activation function | Tanh |
Learning rate | 0.1 |
Momentum rate | 0.5 |
Maximum epochs | 3000 |
Window epoch | 1000 |
Threshold for error | 0.01 |
Random sampling | 50 |
Maximum repeats on stepwise | 10 |
Maximum loops on the whole modelling process | 10 |
Cross-validation | Monte-Carlo with the ratio of 60:20:20 |