Research On Function Discovery Of Gene Expression Programming Embedded With Genetic Algorithm And Spark

With the advancement of society and the development of science and technology,various industries related to IT technology have shown rapid development,and the massive information related to them has also come on stream.How to extract useful information from vast amounts of data and information and help us better Predicting and guiding relevant practical work is a major dilemma currently facing us.Function discovery is one of the important branches of data mining.The purpose of its existence is to deal with massive data,and ultimately find out the existing laws and trends,and form a function model.This article focuses on the application of gene expression programming(GEP)in function discovery.GEP is the third type of new evolutionary algorithm proposed by Ferreira on the basis of inheriting genetic algorithm(GA)and genetic programming(GP).It concentrates on the advantages of GA and GP,and overcomes the inadequacies,eventually achieving the purpose of solving complex problems in real life through simple coding,and is more suitable for application in function discovery.The function discovery problem can be understood in the final analysis as the function with the highest degree of fitting to the original function through the optimization analysis.Among them,the processing method of the numerical constant is one of the important factors that determine whether the GEP algorithm can solve the function discovery problem well.This paper presents a gene expression programming function discovery algorithm embedded with genetic algorithm(GEP-GA).The improvement of this algorithm is to propose a new constant optimization method.In the process of function discovery,the optimization process of each generation is processed through two measures: Firstly,under the premise of a fixed set of constants,the function structure is determined by the conventional GEP algorithm;secondly,Optimizing the constants of the function structure obtained in the first stage by GA and the constants obtained after optimization as a fixed constant for the next generation.The experimental results show that GEP-GA is more effective and more intelligent than the previously specified constant method(GEP-MC).When GEP-GA solves the problem of function discovery,as the problem scale increases and the function complexity increases,GA shows some defects in the global search optimization gradually,for example,the search speed is slow,the evolutionary time is long,the local search ability also becomes worse and so on.How to solve the defects of GA in practical problems is the focus of extensive attention of researchers in recent years.In the face of the above problems,this paper makes full use of the natural parallelism of genetic algorithms and combines the high-speed parallelism of Spark clusters with the natural parallelism of genetic algorithms.The experimental results show that the addition of Spark reduces the possibility of premature convergence of GA,improves the quality of the solution,and shows good results in terms of speeding up searches,reducing evolutionary time,and improving local search capabilities.