Evolutionary Computation - Computational Intelligence in Time Series Forecasting - page 196

1 to the last bit of the same individual. Since the probability of mutation is
generally very low, only a very few bits may undergo mutation of an individual.
If a new individual is generated through mutation then the population counter is
incremented by 1, starting with

u , where po N is the total number of the
population in a particular generation. Otherwise the population counter remains the
same. Using the mutation operator, only 30% of the population, in our case, is
created for next generation (see Figure 5.1).

0.2 N

pop

Generation = 0

Generate the initial
population randomly

Calculate the fitness of the
chromosomes

Output the
best
chromosome

Yes

Terminate the GA run ?

No

Genr = Genr+1

N pop-counter = 0

Yes

N popcounter = N pop ?

No

Chromosome selection for Genetic operation with probability
P(C), P(M), P(R) for crossover, mutation, reproduction resp.

Select one
chromosome
through RW

Select two
chromosomes
through RW

Select one
chromosome
through RW

Produce two new
offspring through
crossover

Produce one new
offspring through
mutation

Produce one new
offspring through
reproduction

Inset two new
offspring into new
population

Inset the new
offspring into new
population

Inset the new
offspring into new
population

N pop-counter = ++2 N pop-counter = ++1 N pop-counter = ++1

Figure 5.1. Structure of the GA implemented for neuro-fuzzy network training