Genetic Algorithms: Artificial Selection vs Natural Selection

Genetic Algorithms (GAs) are a stochastic searching and optimizing method inspired by the biological mechanism of natural selection and evolution. To improve the searching power of GAs for complicated problems, many deterministic measures, particularly, experience and/or expert knowledge-based heuristic rules, have been studied in the existing literature. This paper proposes a potentially more useful general methodology of integrating deterministic strategy with the original stochastic method of GAs. Apart from the mechanism of stochastic natural selection, this paper introduces a more deterministic method of artificial selection as a crucial step to designing the new GAs. An artificial selection gene database contributes a lot to the good performance of the new GA. The concept of artificial selection could even possibly prepare GAs for on-line/real-time implementations in dynamic complex systems. Two case studies – the travelling salesman problem and operation/route planning problem – are conducted and the results prove that the performance of the new GA is significantly improved. Index terms Genetic Algorithms, Natural Selection, Artificial Selection, Optimization.