Improvement Of GEP Algorithm And Its Application In Actual Process Modeling

Gene Expression Programming(GEP)is a new evolutionary computation algorithm developed on the basis of genetic algorithms and genetic programming algorithms.Because of the clear structure of the output solution and the diverse structure of the solution,it has been widely concerned by the academic community.However,it also finds that the GEP algorithm has insufficient local search ability and low output resolution when dealing with some problems.These restrict the development and development of the advantages of the GEP algorithm itself.Therefore,it is of great theoretical and practical significance to study the GEP method,improve its performance and apply it to engineering problems.This paper first gives a brief introduction to the basic principles and development status of GEP.Then,through the analysis of GEP modeling performance,a improved GEP method with stepwise parameter optimization is proposed for the problem of the fitting error of GEP in the evolution process.The method keeps the structure of the function expression represented by the gene in the evolution process of GEP unchanged.By introducing particle swarm algorithm and differential evolution algorithm,the parameters of the function expression are optimized in stages,which effectively improves the performance of GEP.The effectiveness of the proposed improved method is verified by simulation experiments.Furthermore,in order to further improve the performance of GEP,based on the above research,an improved GEP integrated modeling method based on phased parameter optimization is proposed.That is,the sample subset of the original data is obtained by random sampling,and the corresponding sample sub-model is trained by using the improved method proposed above,and the output of the sub-model is taken as the input of the GEP integration method,and the original sample output is re-modeled as the integrated target value.Processing to improve the prediction accuracy of the GEP method,That is,the sample subset of the original data is obtained by random sampling;the corresponding sample sub-model is trained by using the improved method proposed above,and the output of the sub-model is taken as the input of the GEP integration method;The original sample output is re-modeled as an integrated target value to improve the prediction accuracy of the GEP method.Simulation results verify the effectiveness of the method.Finally,the actual method is used to apply the improved method to the modeling of copper concentrate grade prediction and thicker feed concentration soft measurement model in flotation process.The experimental results verify the effectiveness of the proposed method.