The Study About Models Of Short-Term Forecasting To Agricultural Products Market Prices

It is a permanent statement that we use something known to deduce something unbeknown.In fact,we always need to find regulations to forecast the unknown data from the known data.It is of great significance to forecast the short-term changes of apple market price for its importance in farm produce.From a mathematical point of view,we need to deduce mathematical models to forecast the short-term changes of apple market price from the known data.This paper uses various of methods to forecast the short-term changes of apple market price.In the first chapter,the background and sense about forecasting the short-term changes of apple market are elaborated.In the second chapter,we will introduce six models respectively.These models contain hidden periodicity model,mixed autoregressive hidden periodicity model,back propagation(BP)neural network model,genetic algorithm BP neural network model,support vector machine regression prediction model and combined forecasting model.In the third chapter,the above six models are modeled respectively.In fact,hidden periodicity model and mixed autoregressive hidden periodicity model are deduced mainly based on their complex value forms.We use the neural network toolbox provided by MATLAB to deduce the BP neural network model.Genetic algorithm is used to optimize the BP neural network model in the genetic algorithm BP neural network model,that is,BP algorithm will be replaced by genetic algorithm(including selection,cross,mutation,etc).Support vector machine regression prediction model is deduced by adopting software package libsvm3.22.The optimal weighted geometric mean combination forecasting model is made of hidden periodicity model,BP neural network model and support vector machine regression prediction model,which is established by a certain weighting method.The weekly data of this paper is from the Ministry of Commerce forecast website,which ranges May 11,2012 from August 26,2016.The number of data is 224.The results show that the prediction accuracy of the time series model is the worst,the accuracy of the neural network model is the second,the support vector machine regression prediction model and the optimal weighted geometric mean combination forecasting model have the highest accuracy.Be different from the traditional regression analysis method,this article does short-term prediction of apple market price from the mathematical point of view.The advantage is that it is not necessary to know the impact factors of apple’s price,naturally do not need the data of impact factors of apple’s price,which avoids the difficulty of accessing data.Through the short-term prediction of apple market price,the forecasting effect is very impressive,which also shows a fact that they are the same between our methods and the traditional regression analysis.The hidden periodicity model has the advantage of modeling the data with roughly periodic characteristics because it can grasp the trend of data fluctuation.This also reveals that we can transform the known data properly so that we apply desired characteristics to the new data.It is more conducive to master the data changes.The intelligent analysis prediction models are more effective than the time series models.This is not only in the prediction accuracy of the model,but also in the lower requirements of the data.The optimal weighted geometric mean combination forecasting model is based on the hidden periodicity model,BP neural networks model and support vector machine regression model.It focuses on the characteristics of each individual prediction model,the prediction performance is higher than any single prediction model.