In this subsection we mention some results related to the performance of the hybrid GA, especially with respect to the crossover operators used.
When comparing the performance of the two cross-over operators, we initially used equal
values for the other parameters like crossover and mutation probabilities as well as the elitism rate.
It turned out that in those cases, the bit equalizer crossover operator worked generally much better.
Next we performed experiments with the optimized parameter values as mentioned in table II.
As was already mentioned above, both the GA with the two-point order based crossover and the GA with the two-point bit equalizer crossover operator were able to find the optimal solution.
In all cases, we tried the GAs a hundred times using different initializations.
In case of using the GA with the two-point order-based crossover with 50 generations (and other parameter settings as mentioned in Table 2), the optimal solution
was not always found, namely in just 89% of the cases.
However, in case of using that GA with 100 generations, the optimal solution was always found in, on average, almost 31 iterations with a standard deviation of approximately 15,5.
In case of using the the GA with the two-point bit equalizer crossover operator a hundred times with 50 generations, the GA was able to find the optimal solution in all cases using, on average, only 22,2 iterations with a standard deviation of approximately 9,8 iterations.
These numbers show that the GA with the bit equalizer crossover operator is both robust
in finding the optimal solution and fast, although the speed of convergence is somewhat sensitive for the chosen initial population.
The following two figures further illuminate these findings.
Both the progress of the average fitness value (tracking error) of the best solution of each generation is shown and the progress of the fitness value of the best solution of each generation in a single, typical run are shown.
For certain other results found concerning the AEX-index tracking problem, we refer to [1].