Neuro-fuzzy Approach - Computational Intelligence in Time Series Forecasting - page 257

Figure 6.9(f). Finally tuned five Gaussian membership functions plot for fuzzy partition of
input u (scaled/normalized) universe of discourse for Wang data. X -axis (input universe of
discourse, scaled/normalized), Y -axis (degree of membership).
Figure 6.9(g). Finally tuned five Gaussian membership functions plot for fuzzy partition of
input y (scaled/normalized) universe of discourse for Wang data. X -axis (input universe of
discourse - scaled/normalized), Y -axis (degree of membership). Note that fuzzy membership
functions in Figure 6.9(f) and Figure 6.9(g) are largely overlapping. Accuracy and
transparency of the model are expected to be further improved if the similar fuzzy sets (for
e.g. G4 and G3 in Figure 6.9(f) and in Figure 6.9(g) G3 and G5 are highly similar fuzzy
sets) are merged and further fine-tuned using genetic algorithm or evolutionary computation.