Prediction Of Wheat Scab In Anhui Province Based On Multiple Models

Wheat scab is one of the worldwide wheat diseases,it can be affected by many factors and it is still challenging to forecaste the rate.Several prediction models have been proposed for the prediction of wheat scab in Anhui agricultural crop diseases and insect pest early warning platform,but the prediction is not accurate enough.Based on the study of wheat scab,the rate of wheat scab can be predicted by analysising the meteorological data and the spike rate of Fusarium wilt,and the specific research work summarized as follows:(1)Pretreatment of wheat scab dataIn this section,the meteorological factors affecting the occurrence of wheat scab were analyzed.The meteorological data of wheat and the average disease spike rate data of wheat in milk ripening period in Tongcheng of Anhui Province during 2005-2017 years(late March,April and early May)were selected through the database of Anhui agricultural crop diseases and insect pest early warning platform.The selected data above were preprocessed and correlated,and the reasons for selecting grey prediction and BP neural network prediction models were also presented.(2)GM(1,1)model for prediction of wheat scabThe average spike rate of wheat scab was taken as the original sequence of the model,and the model was solved by using MATLAB tool.Because of the fluctuation of the actual value,the precision of grey prediction model was not accurate.(3)BP neural network for prediction of wheat scabFirstly,the meteorological factors were used as input samples of BP neural network,and the spike rate of wheat scab was used as the output sample of the network.And then the training samples were solved by MATLAB tool.Finally,the error was analyzed and the samples were validated by training network test and the average relative error of test data was about 8.19%.The BP neural network model provided higher accuracy,and the prediction result can be guided for gricultural operation.(4)BP neural network optimized based on genetic algorithm for prediction of wheat scabFirstly,we can obtain the optimal initial weights and thresholds through genetic algorithm.Then MATLAB tool was used to train and analyze the optimized BP neural network based on the optimal initial weights and thresholds,and the average relative error of test data was about 6.15%.The combination can stop BP neural network from falling into local optimum.The optimized model can improve the convergence rate and the reliability of the prediction.This paper predicts the occurrence of wheat scab in Anhui by the above three prediction models,and it is proved that the improved BP neural network with genetic algorithm can improve the convergence rate and the reliability of the prediction,and can be used for the prevention and control of wheat scab,the models will provide guidance for agricultural operation.