Font Size: a A A

Combined Forecasting Of Grain Yield In Anhui Province Based On Neural Network

Posted on:2017-12-19Degree:MasterType:Thesis
Country:ChinaCandidate:M Y GaoFull Text:PDF
GTID:2348330488480041Subject:Agriculture
Abstract/Summary:PDF Full Text Request
As everyone knows the "hunger breeds discontentment",in the modern history of China,a large number of people have been killed due to the large-scale famine caused by the problem of food,the population has dropped sharply,and even cause social shock caused by the historical tragedy.Therefore,the food problem relates to the national plan,affecting the survival of a country,is the foundation of national development.2016,from China to achieve the first one hundred years of the goal,that is,in 2020 the overall completion of a well-off society in four years,during which all work of the country are in the efforts towards this goal,which guarantee food production is essential for basic work.Over the years grain yield prediction has been in the field of agriculture is an important work to predict accurate food production to countries in other areas,including GDP,industrial structure of national economy development research.According to the National Bureau of Statistics survey data show that in 2014 the total grain output of Anhui Province,the total grain yield of tons,accounting for the national proportion of 5.63%.single production ranking from 21 to 19,the overall strength has improved,and the annual total output reached a record high of 7 consecutive years.The group has about Anhui province grain yield data tells us that the total grain yield of Anhui Province the accurate analysis and prediction of The overall development of our country and economic construction in both theory and practice has a certain significance.There are many methods to predict the grain yield,mainly including the grey forecasting method,regression forecasting method,principal component analysis method and neural network forecasting method,the neural network and other prediction methods comparison,prediction accuracy is relatively high,its architecture inspired from biological God by the function of the network,trying to mimic the memory of the brain neural network and deal with a large amount of information,generally by neurons as basic unit by different connection mode and different combination of network construct a nonlinear and adaptive information processing system is a commonly used mathematical and statistical methods.In essence,it is the to use function to construct local spatial node A standard learning method of mathematical statistics for a large number of data processing.This paper mainly for grain production in Anhui Province in 1965 to 2014 using neural network prediction.Firstly,it analyzes the model of neural network,learning methods and network structure,and followed by in-depth study of the BP neural network,RBF neural network,GRNN neural network,and based on combination of Iowa operator and three kinds of grain yield prediction method and practice in BP neural network has slow convergence speed,or fails to converge,the initial connection weights,threshold and network selection and random.The value of large or small possible over fitting and non convergence;RBF neural network because of its hidden Containing layer neural network unit of local constraints cannot guarantee the optimal hidden layer to determine the,also hidden layer unit number is usually fixed,often through trial and error experiences,consume a lot of time;GRNN neural network basis function center and width of the neuronal activation degree and influence of neural network function approximation ability,the traditional network learning rule is prone to local minimum convergence results,and even not convergent.Traditional combination forecasting method is weighted assignment for each individual prediction methods of prediction accuracy,and corresponding to each prediction weighted coefficient is consistent with that of the prediction accuracy of the method at every time point.However,the actual prediction process,each prediction method corresponds to the prediction accuracy is not consistent,namely a point prediction accuracy is high,in another point prediction accuracy is low,so the need for new combination forecasting method according to different prediction precision of the different point in time according to the actual situation gives the weighted average coefficient.Based on combination of Iowa operator and prediction model is introduced to Iowa operator to each individual prediction methods and the prediction accuracy in a certain level of order and Weight assignment,obtained for the total error of the squared error and the way to establish prediction model,and design different kinds of model evaluation standard detection model and experimental analysis combination model than the single forecast model prediction effect,and can improve the prediction accuracy and suitable for the prediction of grain production in Anhui Province.
Keywords/Search Tags:BP neural network, GRNN neural network, RBF neural network, IOWA operator combination, grain yield, prediction
PDF Full Text Request
Related items