The Improvement Of Gene Expression Programming And Its Application In Knowledge Discovery Research

Gene Expression Programming algorithm(GEP) is a new type of intelligent evolutionary algorithm which can process high latitude and uncertainty factors. It can dig out the knowledge hidden in the data, such as rules, models and so on, and does not require any prior knowledge. The algorithm with unique way of coding, good ability of data mining and the processing capacity of highly nonlinear system has won numerous researchers' attention, and has been widely used in various fields.The main research works in this thesis are that improving the standard GEP algorithm and applying it to the two great problems of Knowledge Discovery, namely wheat aphid population model and building engineering cost prediction, and which could provide the basis for the prediction of wheat aphid population, the feasibility study of project and reasonable design scheme. The concrete works are as follows:(1) On the basis of reading a lot of related literature and the methods of prediction, this study summaries the important knowledge of applications, introduces the process of wheat aphid population modeling and the extraction and classification of the construction project cost prediction features, clearing the analysis objects; sums up the principle and encoding of the GEP and analysis the process and the basic operation of the algorithm in detail.(2) Based on the theory of artificial intervention, this thesis proposes a Double system co-evolutionary gene expression programming(DSCE-GEP) which consists of artificial intervention system and natural evolution system. The former includes individual intervention and population intervention operations. Individual intervention mainly depends on the gene pool with the high quality to execute the increase the optimal and rouging operations, in order to improve the quality of individuals; while the population intervention operations uses the information entropy to improve the population diversity through introducing some random individuals and mirror individuals.(3) Aim at the improved algorithm above, this research validates the effectiveness and progressiveness of DSCE-GEP through the comparative simulation experiments. Then, this article applies the algorithm to wheat aphid data provided by the China Academy of Agricultural Sciences to build model, and also uses the engineering data listed in literature to model and forecast. The experiment results show that, the effect of the wheat aphid population model and building engineering cost prediction model based on gene expression programming in this thesis is superior and the prediction accuracy is higher.