The Analysis And Solution For The Limitations Of GEP Decoding

Gene Expression Programming (GEP) is one of the newest member of evolutionary algorithm, which combines the advantages of both Genetic Algorithm (GA) and Genetic Programming (GP). It has an easy operation and a simple encode. It also has a flexibly expressing capability and a strong searching ability. Its evolutionary efficiency is 100-60000 times higher than genetic programming system and the superior performance thus has attracted a growing number of researchers. Lots of improvements of GEP, such as initial population, chromosome coding, genetic operation, have been achieved. And also, the combinations with traditional artificial intelligence methods make better evolution efficiency. Researches on Pattern Law and convergence of GEP in theory have also yielded some outputs.GEP is now widely used in symbolic regression, classification, time series prediction. After all, since the invention of the GEP algorithm there are only about ten years. Many fields in both of theories and applications of GEP are waiting for further explorations.Based on the work of previous researchers, this paper studies the research status, principles, decoding methods and applications of GEP. The work mainly includes five parts as follows:The first part describes the theories and technologies of the traditional genetic algorithms and genetic programming.The second part introduces the key factors and processes of the gene expression programming and analyzes the essential differences among GEP, GA and GP.The third part analyzes the deficiencies of traditional GEP decoding methods and proposes a new GEP decoding method namely the Non-physicaltree decoding, which establishes and operates the expression tree in the decoding process without a real physical sense and thus can reduce computation complexity;In the fourth part, the GEP algorithm is applied to the prediction of the stock trend and symbolic regression and achieves good results;The last of the text summarized the paper , analyzed the problems of this study and take a good prospect of future development.