It had been learnt from [4] that, increase in accuracy
alone may not however be the only justification
of including this particular factor into the model, as
there existed risk of over fitting. To clarify the decision,
the accuracy predicted in each year was averaged
over the models with 8 and 9 factors. The MLP
did better than RBF with the accuracies of 71.3, 78.1
and 80.85 percents compared to 64.8, 74.45 and 81.05
percents (Fig 11).