Improvement of gene expression programming algorithm
To solve the problems of gene expression programming (GEP),such as slow convergence rate,premature convergence and easiness to fall into local extreme points,three improved methods were proposed.The adaptive evolutionary parameter was designed.According to the number of generations and the ranking of fitness value in the population,the recombination rate and mutation rate were dynamically adjusted.To further expand the search space and avoid premature,the population was stratified by their age.GEP was ported to the Spark framework for distributed parallel computing,so that the algorithm dealt with a large number of search tasks in a relatively short period of time.Experimental results show that,compared with the traditional GEP,the improved algorithm has higher convergence speed,prediction accuracy and stability.
gene expression programmingadaptive evolutionlayered population structureSpark distributed computingprediction
NETL
NSTL
万方数据
维普
