All selected models show very low MSE values (ranging
from 2.8 103 to 107) and values of R2 close to 1.00 in validation
process. This indicates strongly that these models
predict data with high accuracy. Certainly we can conclude
that both methodology and presented results allow a proper
ANN model selection with a statistical basis. The above
mentioned selection process was developed in a Matlab's
script file available upon request to the authors