The prediction results obtained with the proposed model
exceed those obtained with the simple neural network. This proves
that it is possible to improve performance by using additional
information from the existing nonlinear relationships between the
concentration of the pollutants and the meteorological variables.
Moreover this information reduces the necessary input data to
make the prediction as shown in Tables 2 and 3, where we can see
that, using MLP model, NA station requires 7 days of previous
information to make the prediction while the proposed model only
requires one day of previous data. For the DIF station, the MLP
model requires 4 days of previous information while the proposed
model requires only one day of previous data.