A New GEP Algorithm Based On Multi-phenotype Chromosome And Its Application

Evolutionary algorithms are a set of statistical optimization algorithms. They are a kind of optimization algorithms which are enlightened by the natural evolution process, especially by the adaptability that individual organisms show in their environment. In the beginning, evolutionary computing has three branches:Genetic Algorithms (GA), Evolutionary Programming (EP) and Evolution Strategy (ES). In the early 20th century 90s, another branch was derived from GA, called Genetic Programming (GP). In 2001, Ferreira presented a new evolutionary algorithm called Gene Expression Programming (GEP) which integrates the features of GAs and GPs.Gene Expression Programming is a new evolutionary algorithm with linear genotype/nonlinear phenotype. Like GAs and GPs, It uses populations of individuals, selects them according to fitness, and introduces genetic variation using one or more genetic operators. The difference in GEP is that in GEP the individuals are encoded as linear strings of fixed length (the genome or chromosomes) which are afterwards expressed as nonlinear entities of different sizes and shapes, called Expression Trees.This paper presents an advanced GEP that based on multi-phenotype chromosomes, called MPC-GEP. The algorithm expresses an individual into several expression trees, thus an individual contains several possible solutions. In this way, the probability of populations in MPC-GEP contain the optimal solution is increased. Experiments on the new algorithm against traditional GEP algorithm are conducted on several benchmark datasets. Results show that MPC-GEP outperforms traditional GEP in function finding in terms of speed.Classification is an important issue in data mining. Nowadays, there are many algorithms applied into classification rule mining, including algorithms based on decision tree model, neural network model and rough set model. Chi Zhou and his colleagues have researched the classification algorithm based on GEP, and got good results. In this paper, MPC-GEP is applied in classification rule mining, and the experiments which then compared with the experiments by Chi Zhou are conducted on several data sets from UCI machine learning repository. Results show that MPC-GEP can get the accurate classification rule in less time.