Study On Multi-expression Programming And Its Application Of Evolutionary Modeling

In scientific research and engineering technology fields, there are a large number of issues need to establish a corresponding mathematical model to describe input-output relationship of abstract systems. However, it is impossible to get a mathematic model which can show the inherent laws of the complex systems' observed data.Evolutionary modeling is a modeling method to search for the optimal model, bases on evolutionary algorithm. It can automatically mine and establish the relation between input variables and output variables. Because of its population (means group of features) way to organize search, MEP can simultaneously search the solution space in multiple areas. And the way that organizing search with population makes evolution algorithm especially suitable for large-scale parallel computing.Now the popular algorithms for the evolutionary modeling are GP and GEE. MEP (Multi-expression programming) is another new algorithm in the field of genetic programming in recent years. MEP has a high ability to searching and mining. MEP shows a stronger modeling capability as compares with other genetic programming, so apply evolutionary algorithm to solve the problem of modeling the MEP is a beneficial attempt.This paper discuses the evolutionary modeling problem applying Multi-expression programming. It mainly divides into three parts. The first part is for the research of MEP algorithm. It introduces the GP developing process:from non-linear GP (Genetic Programming) to linear GP (general GP, GEP, and MEP) and describes the basic production process MEP algorithm. It focuses on the basic algorithm of the key technologies of the MEP encoding, fitness function, genetic methods and structure of the basic MEP algorithm. The second part is for evolutionary modeling bases on the basic MEP algorithm. It researches the basic principles of evolutionary modeling applying the basic MEP algorithm, and through modeling example shows the basic MEP algorithm in evolutionary modeling viability. The third part is for evolutionary modeling bases on the improved MEP algorithm. In order to improve the fit and precision of the basic MEP algorithm, a new improved MEP algorithm is proposed. It introduces niche technology, and through modeling example shows the MEP algorithm to improve the rationality and effectiveness. It makes a conclusion that this algorithm can generate a function model having a better fitting precise.Main work and innovation of this article are as follow: 1. We study the basic principles of evolutionary modeling applying the basic MEP algorithm, and through modeling example shows the basic MEP algorithm in evolutionary modeling viability.2. We propose the improvement MEP algorithm bases on niche technology. And by modeling example shows the MEP algorithm to improve the rationality and effectiveness. We find through compare the analysis of the results of the models that the models have a better fitting precise and can more reflect the connections of the complex system data.