A Self-adaptive Genetic Algorithm Based on the Principle of Searching for Things

Abstract

This paper proposes a new self-adaptive genetic algorithm。This new algorithm divides the whole evolution process into three stages. At each stage, the new algorithm adopts different operation method. The main ideas are grading balance selection, continuous crossover operation. The new algorithm designs especially self-adaptive mutation probability according to the principle of searching for things. Numerical experiments show that the new algorithm is more effective than the comparative algorithm in realizing the high convergence precision, reducing the convergence generation and good at keeping the stability of the adaptive genetic algorithm.