Statistical analysis of genetic algorithms and inference about optimal factors

摘要: Abstract In this paper, statistical analysis, in the form of regression models and fractionalfactorial experiments, has been applied to genetic algorithms (GA) – a search methodfor non-linear constrained function optimisation. The distribution experimentalresults way make normal have established. Then, GAfactors whose significance is invariant environmental noise beendetermined. Finally, describing two characteristics searchefficiency terms significant GA factors obtained. These modelsenabled calculation optimal values for factors. After substitutionof these values, algorithm performance substantially improved.Keywords: models, Taguchi method, fractional factorial design, Weibullmodel, algorithms, optimisation control. 1. Introduction Genetic are an tool that computationally emulates process ofevolution. They number engineering biomedical problems whichrequire optimum through large space candidate solutions (Mitchell,1996).Genetic combine elements stochastic directed strategies (Michalewicz,1992). Stochastic features present themselves information gatheringand exchange implemented by mutation crossover operators. operators play anextremely important role with respect speed robustness themethod.Unfortunately, there no simple analytical expression or theoretical model describes theperformance method’s One estimate effects ofsignificant use methods analysis inference. This majortask report.1.1. Exploration vs. exploitationOne areas where shown their effectiveness wide class ofproblems require implementation efficient search. able toperform multi-directional using population explores thesearch (Michalewicz, 1992).