The Improvement Research On Evolution Strategy Of Genetic Algorithm

Genetic algorithm is a random search algorithm of optimization algorithms,it is an intelligent algorithm oriented by a fitness function.It has a lot of advantages in practical application through learning from the rule of evolution of biosphere.People has make a thorough research on its theory and application and has make a great achievement since the birth of genetic algorithm.People has also make many improvements to the genetic algorithm research according to its own deficiencies and disadvantages,thus the performance of genetic algorithm is also improved to some extent.In this paper,we have make intensive researches on the evolution strategy,crossover probability values and encoding methods through the operational mechanism of genetic algorithm.The basic theory and operation of genetic algorithm was introduced in the first place,some researches was made about the operations of binary genetic algorithm and real-coded genetic algorithm respectively.We have make some improvements on the evolution strategy and crossover probability values of genetic algorithm,and program repectively using binary and real number.Some problems were found as follows:(1)The evolutionary strategy of genetic algorithm in the existing literature is generally "preserving elite individuals in parent generation-selection-crossover-rutation-replace elite individuals".(2)The crossover probability values had not been determined until now,so how to determine the value is the problem to be solved. In this way of genetic operation, even excellent individuals produced in the crossover operation process may also become bad in the process of mutation. So how to preserve produced excellent individuals needs to be improved for existing evolutionary strategy;In this paper,some typical sonsiderable complexity functions are selected and have simulated tests on them using binary code and real code respectively,analyzing and comparing the results of the tests.Thus we can prove the improved algorithm can improve the performance of the algorithm proposed in the paper.In this paper,the main improvements of genetic algorithm and the results are as follows:(1)The elite reserved strategy of the improved genetic algorithm not only retains the outstanding individuals of parent population,but the elites retained after the crossover proposed in this paper.In this way,the destroyment of excellent individuals after the crossover can be prevented in mutation,so as to accelerate the speed of convergence to the optimal solution.The mean algebra to the optimal solution is not only reduced,but the average time for optimal solution is reduced,thus the performance of the genetic algorithm has been improved. (2)The crossover probability value should be1is proposed.We regard the value of crossover probability as one of the research objects,analyzing the effect of the value of crossover probability on the performance of genetic algorithm,and we proposed that the crossover probability value should be1.We have tested the functions chosen in different values as0.5,0.6,0.7,0.8,0.9,1.The testing results show that the bigger the crossover probability,the faster the speed,and the less the average run algebra. Verifing the scientific and feasibility of the crossover probability value1proposed in this paper.In the mean time,the deficiencies of the determing difficulty of the crossover probability and the lack of theoretical basis in determing the crossover probability can be overcomed.(3)In this paper,the genetic algorithm is improved through in-depth research on it,and adopt the method of combining theoretical analyses and practical applications,Apply binary genetic algorithm and the real genetic algorithm to functions optimization unconstrained and constrained,testing the validity of the improved algorithm proposed in the paper.In conclusion,summarizing the research works that we did in this paper,the prospect is made.