Abstract

A genetic Algorithm is an evolutionary algorithm that models and simulates biological behavior, whether evolution or genetics, to reach a high-quality solution for search and optimization problems. There are many areas and applications to which genetic algorithms can be applied, like machine learning, feature selection, engineering design, and function optimization. Three leading operators must be applied to each generation's reproduction process; the first is the Selection process, which is applied to the initial population to select the candidate parents to mate and recombine to produce the next generation(offspring). The second operator is a crossover, which is applied to the selected parents from the previous operation (Selection) to make new individuals (offspring) carrying the same traits from parents by combining the parent's chromosomes; the last operator is a mutation, which is applied to the new offspring after crossover. Mutation operation aims to change the value of the chromosome gene randomly. In this research, the selection process will be demonstrated in detail. Then, fitness proportionate selection (FPS) will be presented as one of the most popular methods used in the selection process. The main problem of FPS is the candidate parent, which will mate and recombine to reproduce the next generation; in some cases, a strong individual can mate with a weak one and produce offspring with lower quality traits than the strong parents as a consequence of trait exchange, which happens between that pair. The researcher proposed an enhancement of the FPS algorithm to ensure that strong parents will mate and reproduce strong offspring and propagate their strong traits to the next generations; the proposed enhancement can be summarized as adding a step to the standards FPS to sort the selected individual in ascending or descending order after selection process and before applying cross over and mutation phases. The researcher conducted three experiments to prove the improvements in the fitness value as a consequence of applying that additional step in the selection algorithm; the experiments were performed with three different population sizes and reproduced 100 generations. The fitness score was measured in each generation, and the researcher presented the fitness score evolution over the GA iterations. The results were precise, proving that the sorted individuals after Selection gave better fitness scores than those obtained by applying the standard FPS.