Gene Expression Programming Based on Mixed Selection and Dynamic Variation

Gene expression programming(GEP), which is a new type genetic algorithm based on genotype and phenotype, is now widely applied in function finding, time sequence prediction and classification.However, due to strong intensity in preferred roulette selection of individuals, traditional GEP often leads to reduction in the diversity of individuals and causes inbreeding;on the other hand, traditional GEP can not adapt to the evolution result of each generation by adjusting variation probability, so that the performance of GEP is affected.In order to overcome the drawbacks above, this paper makes two amendments to the traditional GEP:The author adopts the mixed selection strategy to maintain individual's diversity and avoid the phenomenon of inbreeding.Dynamic variation is introduced to reserve the highfitness gene segments and eliminate the low-fitness ones by adjusting variation probability of individuals according to the evaluation of individuals' performance.Through experiments, this paper verified the validity of two amendments.