The performance profiles with different training epochs for test and validation sets. The ANNs were trained with different numbers of hidden layer nodes as follows: (a) 16 hidden nodes, and (b) 22 hidden nodes. The profiles show that the optimal performance for the test and validation sets did not occur at the maximum training epochs. RMSE became more stable and relatively lower when the training epoch number was between 350 and 500. The best observed error for the test set is RMSE 0.0767 at 450 (b).