Average Concept of Crossover Operator in Real Coded Genetic Algorithm
Rosshairy Abd Rahman, Razamin Ramli · 2013
As the most important search operator in a Genetic Algorithm (GA) approach, many procedures have been proposed to accomplish the idea of a crossover. As a result, knowledge in crossover has incorporated special features such as statistical elements (i.e. arithmetic crossover) and natural observation (i.e. queen bee crossover) to name a few. Thus, this paper proposed a mean or average concept of crossover for fitter parents to produce a new offspring in a GA based approach in an animal diet formulation problem. Experiments using real data were carried out involving GA models with average crossover and one-point crossover. Subsequently, the incorporation of power heuristics as a repair operator was investigated to find the best combination of ingredients, while removing the unwanted ones. Comparisons were made between GA models incorporating repair operator with different crossovers: average crossover and one point crossover. The results show that the performance of average crossover is comparable with that of the one- point crossover. The inclusion of the repair operator provides an advantage that shows interesting solution for the tested problem.