Genetic Algorithm Improvements And Applications

All researches in this dissertation have systematically studied the genetic algorithms(GA) being the intense popular interests in the international academia at present.based on the analysis and research for coding and operator of genetic algorithms,the dissertation has put forward the objective warrant for multi-parameter coding,discoverd and expounded the hypostasis of crossover operators.brought forward a plan to settle non-completeness of one-point crossover operator and devised a kind of feasible and impactful crossover operator, moreover validated and compared them through two instances. Considered the above problems of genetic algorithms,the main contributions are presented as follows.(1)Aiming at the multi-parameter coding problems which manner should be adopted-series or crossover, the dissertation found objective gists for coding mode adopted. In general ,the tactics or ways of genetic operations have great impacts upon the function of crossover operation.In many instances,Once the mode of coding was selected,crossover operation was determined too.Therefor,the coding techniques should be thought over before applying genetic algorithm.In this dissertation after multi-parameter coding was analysized by software SAS,the randomicity of multi-parameter mapping coding was solved and an gist of objective judgement was also given .(2)Based on the comparison between floating point numbers and binary system, the commoner concousions were thought out, it offers theoretical guidance for devising new coding mode and crossover operators .(3)The kernel operation of genetic algorithms is crossover operator,by the comparison between one-point , two-poin, multi-point and uniform operators,the essence of crossover operator was discovered,and laid a sound foundation for devising highly efficient crossover operators.(4) As non-completeness occurs in the crossover process,the dissertation brought forward the notions 'dead-code' etc.devised a kind of crossover operators cal led 'modular-two-add' . It can not produce the 'relative dead-code' in crossover process, its computing quantum does not increase,although the convergence speed increases.At the same time it also brought forward the improvement of one-point crossover operator,and combined two examples to validate effect of operation between the improved operator and the former operator.