Application Research Of Evolutionary Modelling Method Based On Gene Expression Programming

In view of the complex problems in life,in order to explore the law of facts and to mine the value of data,people hope to construct a mathematical model which can reflect the factual situation and the future trend of facts by a relatively perfect mathematical method.Gene expression programming(GEP)is a mature evolutionary algorithm,which can automatically build models and find functions which can represent the interior rules in complex data by its self-organization,self-adaptation and self-learning characteristics.The thesis focuses on the historical data of complex dynamic system,uses GEP algoriyhm to construct a prediction model and uses the model to describe the dynamic law of time series data,so as to grasp the future development direction of events.The thesis proposes modification strategy and optimization methods aiming at the shortcomings of the traditional GEP algorithm in dealing with problems and the different use scenarios of the model.The experimental results demonstrate the effectiveness of the work.The main work of this thesis is as follows:1.Take advantage of GEP's advantages in automatic modeling of complex functions and the ability to solve real-world problems,this thesis propose a hybrid highway toll prediction model based on GEP to excavate the rules of variation in highway toll.In addition,since the toll reduction policy will cause the low fitting efficiency as well as the inaccuracy of capturing extreme point among special months,this thesis proposes a modified algorithm based on central balance(CB-GEP).By collecting the historical data on the toll revenue of 12 companies,the results fully verify the effectiveness and accuracy of the proposed algorithm.2.An evolutionary modeling algorithm for high order ordinary differential equations based on GEP has been proposed.In view of the limitations of low-order ordinary differential equations in complex dynamic systems,which are difficult to accurately represent the inherent laws and development trends of things.In this thesis,a high order ordinary differential equation based on GEP(GEP-HODE)is constructed by utilizing the ability of high order ordinary differential equation to describe complex dynamic systems and the modeling advantages of GEP in function discovery.The GEP-HODE is verified by predicting the trend of stock price;the results show that compared with the traditional forecasting model,the algorithm has higher forecasting accuracy for the trend of stock price.3.A parameter optimization method based on Krill Herd algorithm(KH)has been proposed to optimize those parameters of GEP-HODE.Since GEP is easy to fall into local optimum and has slow convergent speed in the later stage of evolution,this thesis adds a modified KH which is based on dynamic pressure control operator(DPCKH)to optimize the parameters of GEP-HODE.The optimization ability of the algorithm is improved by giving different inducement effects to excellent individuals in different periods.The experimental results show that the optimized models have higher prediction accuracy.