Application Research Of The Gene Expression Programming In The Evolutionary Modeling

Along with the development of the technology and society, a growing number of practical issues need to be solved by the use of mathematical methods in in science and engineering applications. As an important method in the system problem research, often the complex practical problem and phenomena are simplified and abstracted to a simple problem through modeling. As a result a simple model can be built to study. However, it is almost impossible to achieve a mathematical model which can accurately reflect the data between the inherent law by the use of the traditional manual method for the observational data of the complex system.The study of the evolution modeling problem based on GEP is focused on in this paper. The main analysis and research is done from the following aspects: Firstly, basic principles of evolutionary modeling are described in this paper and the key technologies of GEP are focused on. Then the coding advantage of GEP is proposed and the performance of GEP and GP both evolutionary modeling algorithm used is compared.Secondly, the basic principle of GEP evolutionary modeling is carried out. After that two modeling examples based on one-dimensional and multidimensional parameter space is operated and the result is compared with the result got from GP. According to the modeling results, the conclusion that there is higher fitting accuracy of the function model generated by the GEP evolution model is verified.Finally improvements for the problem that gene drift and convergence speed delayed appear in the process of gene mutation is carried out:drift suppression operator is added and the mutation radius is adjusted dynamically in the mutation process. Afterwards gas emission modeling is built based on the improved GEP. Compared with the modeling results of traditional GEP and GP algorithm, the conclusions that a shorter average time costed to obtain the model successfully based on improved GEP, higher fitting precision and more accurately reflecting the inherent link and laws of the complex system data are obtained.