An Improved Genetic Algorithm And Its Application In Function Optimization

Genetic algorithm,which is presented by Holland in 1975, is a kind of evolution algorithm. The algorithm is widely used for its little auxiliary information, easily integration with other knowledge, simple steps and running without any complex process.Meanwhile,the algorithm also is adaptive and parallel. Genetic algorithm is good at solving some complex nonlinear optimizations which are difficult for traditional optimization algorithms, but the disadvantage of slow convergence and premature phenomenon is hard to handle.In order to overcome the shortages of genetic algorithm, lots of improvements are proposed. With the help of some researchers, the crossover probability and mutation probability have great influence on genetic algorithm's performance. Therefor the new algorithm which the crossover probability and mutation probability are adaptive is proposed. By adjusting automatically, the new algorithm is hoped to have a excellent performance with balanceing global optimization and convergence rate.Summarizing former research, the adaptive genetic algorithm is analysised. The recent adaptive genetic algorithms are restricted to original adaptive algorithm's idea. The fitness of chromosomes which is less than average fitness of population are disrupted by higher probability, so the disrupted chromosomes' knowledge are not rationally used. According to the defect of that, a new genetic algorithm is proposed to adaptively adjust the chromosomes' crossover probability and mutation probability. By adaptively adjusting the disrupted chromosomes' probability, a good performance is hoped to achieve.In order to test the new algorithm's ability, the new algorithm and other adaptive genetic algorithm are applied in function optimization. The experiment shows that the proposed algorithm is clearly improved in global optimization and stability.