Controlling The Problem Of Bloating Using Stepwise Crossover And Double Mutation Technique

Abstract

During the evolution of solutions using geneticprogramming (GP) there is generally an increase inaverage tree size without a corresponding increase infitness---a phenomenon commonly referred to as bloat.The conception of bloat in Genetic Programming is awell naturalised phenomenon characterised byvariable-length genomes gradually maturating in sizeduring evolution. 'In a very real sense, bloating makesgenetic programming a race against time, to find thebest solution possible before bloat puts an effectivestop to the search.' In this paper we are proposing aStepwise crossover and double mutation operation inorder to reduce the bloat. In this especial crossoveroperation we are using local elitism replacement incombination with depth limit and size of the trees toreduce the problem of bloat substantially withoutcompromising the performance. The use of local elitismin crossover and mutation increases the accuracy of theoperation and also reduces the problem of bloat andfurther improves the performance. To shew our approachwe have designed a Multiclass Classifier using GP bytaking few benchmark datasets.