Long-Term Rainfall Information Forecast by Utilizing Constrained Amount of Observation through Artificial Neural Network Approach
Table 3
ANN parameters and configuration for future rainfall forecasting.
Parameter
Configuration
Network configuration
Input (5); hidden (6) and hidden (4); output (1)
Transfer function
Linear transfer function (purelin) at hidden 1, the hyperbolic tangent sigmoid transfer function (tansig) at hidden 2, linear transfer function (purelin) at the output layers
Training algorithm
Gradient descent algorithm (traingda), the adaptive learning rule
Learning rate
0.05
The ratio for the learning rate increment
1.05
Target performance in terms of MSE
0.002
The optimum number of epochs to train
30
The training performance range of eleven networks among years 2010–2020