The Research Of Evolutionary Performance To Multiple Value Coding Genetic Algorithms

Coding schemes is very important to applications of genetic algorithms. From the size of code alphabet, it can be divided into two value coding and multiple value coding genetic algorithms. Comparing with two value coding, multi-value coding genetic algorithms represents different evolutionary performance. With the solving problems becoming more and more complicate, advantages are manifested through using it. It has the value in theory research and practical application to study the evolutionary performance of it.In this paper, aimed at the particularity of problems as a result of multiple values, through schema gene distilling and emerging lacked allele, methods of improving the capability of multi-value coding genetic algorithms were discussed. Three aspects have been mainly included.Firstly, comparing evolutionary performance with two value coding genetic algorithms, multi-value coding genetic algorithms has better capability in the stability and searching precision.Secondly, with the evolution iteration, a part of genes have an exponent increase. Because these genes are not genes of the optimal result, algorithm is got into local searching and is hard to affirm needful genes to let algorithm get the global solution. Based on the features of multi-value chromosomes in population, schema genes distilling method is shaped, then a method is proposed by using these genes. It can promote statistic schema to converting into the schema of approximate optimal solution and the global searching performance of population can be improved effectively.Thirdly, allele lack is an inevitable problem. Through analyzing the difference of this with two value coding genetic algorithms, summarizing the outlines to resolve it in multi-value coding genetic algorithms, a method of emergence and survival of lacked allele is come up with. It can strengthen recombination function with emergence of lacked allele.The experimental object is the multiple choice knapsack problem(MCKP) [1]. MCKP is a large combination optimization problem. It can be used in many fields. Such as capital budget, large product structure optimization problem[1,2], and so on. The validity is showed by applying these methods to the genetic algorithms to solve the multiple choice knapsack problem.