Forecasting And Application Of Grain Price Based On Multi-scale Combined Model

Grain is not only an important strategic material which have relationship to the people’s livelihood and national economic security, but also be people’s most basic living materials. Over the past 12 years, China has achieved a continuous increase in grain output, but it is facing the current situation of "shortage of demand", as well as the gap is increasing. The proportion of grain imports in China has increased continuously in recent years. Volatility in international grain prices overall sharply higher, China is lack of international grain pricing power as same time. It has a direct impact on the national grain security and economic security because it passively accepts the substantial increase in international prices. It is very important significant to predict the price fluctuation and trend effectively on the national’s economic policy formulation, enterprises’ decision making and production planning of the farmers.This paper selected wheat, rice and corn as the representatives to research seasonal fluctuations based on X-13A-S and studied the influence of trend cycle factors and irregular factors on the price. The results showed that price of grain had obvious seasonal fluctuation, which was consistent with their respective growth cycle; short-term changes in food prices was jointly determined by the rules of factors, trend and seasonal factors, the irregular factors, the greatest impact, and prices in the long-term change was determined by the trend factor.Respectively used spectrum analysis and wavelet analysis to research periodicity of food prices. It determined the food cycle of existence and cycle length. Finally, the reason of food price cycle were analyzed. Prices had a period of 2-4 years, which was related to the adjustment grain production of the farmers according to the price. Under the combined effects of different factors and the continuous development of financial, grain price instability was more and more strong, the next cycle of grain would had a trend of more and more short.This paper constructed a new multiscale combination forecasting model based on the idea of decomposition-reconstruction-integration, it used the ensemble empirical mode decomposition (EEMD), grey relational analysis, artificial neural network (ANN), support vector machine (SVM) and so on. The specific process of model construction was:firstly used EEMD method to decompose the grain price series; then calculates the gray correlation coefficient of the decomposition, while considered the frequency fluctuation would be the dividing refactoring; thirdly, selected different pre measurement method to predict different fluctuation of reconstructions; finally, selected SVM method to integrate prediction results.The prediction and analysis of wheat, rice and maize were carried out by using the multi-scale combined model constructed in this paper. The empirical results showed that single model constructed in this paper the multiscale combination forecasting model was better than that of GM (1,1), BP neural network, support vector machine (SVM), ARIMA model as well as be better than arima-svm combination model and that based on EMD and EEMD other multiscale combination model. At the same time, it was found that the accuracy of the price forecasting of the three kinds of grain is very close, which indicates that the multi-scale combination model constructed in this paper is suitable for the prediction of grain price.