Fuzzy Clustering in Fitness Estimation Models for Genetic Algorithms and Applications

In complex situations, genetic algorithms need a large number of fitness evaluations before satisfactory results are obtained. In many real-world applications fitness evaluation procedures ca be computationally costly. Often, actual decision-making circumstances demand solutions as fast as possible, requiring from genetic algorithms good solutions within short periods of processing time. This paper suggests the use of fitness estimation models based on fuzzy clustering as a means to improve genetic algorithms performance in complex problems. The aims are to decrease the computational effort required to evaluate individuals using fitness estimation models, to decrease genetic operations complexities, and to keep solution quality. The fitness estimation models suggested in this paper perform well in classic benchmark problems and an actual train scheduling problem for a single-track freight railroad.