A Fuzzy Taguchi Controller to Improve Genetic Algorithm Parameter Selection

The selection of operators and parameters for genetic algorithms (GA) depends upon the situation, and the choice is usually left to the users. Identifying the optimum selection is very time consuming and, therefore, it is important to develop a system which can assist the users in their selections. In our fuzzy Taguchi controller, we present a hybrid system, which combines the Taguchi method with fuzzy logic, to select near optimum settings for the design parameters. The Taguchi method selects an optimal orthogonal array from experimental design theory, to reduce the number of experiments required to study the parameter space. Our controller uses this array to determine the selection for fuzzy membership in the dynamic selection process. It then applies fuzzy logic to evaluate the beneficial genes which affect the GA performance. We use the hybrid procedure to produce evidence from simulations and this information is then used to refine the GA behaviour. The system utilises a fuzzy matrix to rearrange the sequence of gene groups within the chromosome and applies a fuzzy knowledge base to tune the GA parameter selection. This provides a simple and easy method to assist users to direct their search and optimisation in an efficient way.