The determination of invariant parameters and operators of the traditional genetic algorithm using an adaptive genetic algorithm generator

This dissertation identified the operators and parameter settings of a Traditional Genetic Algorithm (TGA) that are invariant across problems as well as those that are problem specific. This Identification was done by examining superior TGAs generated with an Adaptive Genetic Algorithm Generator (AGAG). This research also designed and developed AGAG, which uses a Genetic Algorithm based on Migration and Artificial Selection (GAMAS) as a metalevel genetic algorithm (GA) to determine the best design of a TGA for a specific environment (optimization problem). This system incorporates ali the TGA components (operators and parameter settings) that have been defined and developed for binary encoded function optimization in the literature. Using AGAG, the best combination of these components were identified that allowed the adapted TGA to perform well in its designated environment. AGAG was used as a research tool to optimize the TGA in the most common test bed of functions for genetic algorithms. The primary goal of this dissertation was the Identification of the operators and parameter settings that are invariant across problems as well as those that are problem specific. This Identification was done by examination and comparison of the adapted TGAs generated for ali test problems. This Identification will help researchers direct their studies to further advance the theory of GAs. The secondary goal of this dissertation was the development of the research tool, AGAG, used in this research, and will be used in future research. Using this tool, researchers and practitioners can develop a TGA that, given any optimization problem, will perform well because its operators and parameters have been determined through an adaptive process.