Improved Gene Expression Programming And Its Application In The Prediction Of The Population Density Of The Aphid

Sitobion avenae is one of the major pests which are harm to the wheat in China.The exact prediction of the occurrence quality of the sitobion avenae is the premise ofcorrect control decision-making and the reduction for the use of pesticides. At present,the predictions of the sitobion avenae are in terms of levels rather than the occurrencequantity. In addition, the ways of predictions are unitary and with many errors. It’s hardto predict in the complex environment, and it is also difficult to predict if there are manyinfluencing factors. It cannot meet the needs of control decisions. The evolutionarymodeling can take the advantage of evolutionary algorithm and some search strategiesto automatically generate the function model with the high fitting accuracy throughrepeated trials. In the present thesis, it firstly introduces a kind of evolutionaryalgorithm: gene expression programming. Following is the concrete analysis of itsprocedure and the realization method. Through systematic analysis, we find that thenumeric constants affect the algorithm of GEP a lot. The processing method of numericconstant is a little bit complex and the effect is not very ideal. Therefore, in the presentthesis, we put forward an improved algorithm of GEP, it optimize the research processof numeric constant algorithm. Besides, it verifies the effectiveness of the algorithm byexperiments. Finally, we adopt the improved algorithm of GEP in the structure ofsitobion avenae’s prediction model. Compare the traditional algorithm of GEP and theimproved one and get the ideal results. The main content of the present thesis are asfollows:(1) To study the basic principle and algorithm flow of gene expressionprogramming algorithm. It introduces the encoded mode, genetic operator and fitness function of the algorithm of GEP in details. And to summarize the characteristics ofGEP and the differences between other genetic algorithm GA and GP.(2) Because of the shortcomings of the numerical constants in the originalalgorithm, we put forward an improved algorithm of GEP. The biggest feature of thismethod is to divide every generation of discovery process into two phases: The firstphase: the standard algorithm of GEP and constant set determine the function ofstructure; the second phrase: To optimize the function of structure constants which areobtained by the first stage through the method of differential evolution algorithm. Whencompared the improved algorithm of GEP in the important literature, the new algorithmof GEP can restrain the algorithm fall into the locally optimal solution, and have agreater probability search for globally optimal solution.(3) To predict the occurrence quantity of the sitobion avenae with the method oftraditional algorithm of GEP and the improved one respectively, and test the model withthe sample data. The experiment declares that the improved algorithm of GEP predictsthe occurrence quantity of the sitobion avenae with higher accuracy. It is of great value.