The Comparison Research Of The Forecast Model On China Shouguang Vegetable Price Index

Recently, as the continuous development and improvement of the vegetable industryin China, we give birth to the first vegetables price index system, China Shouguangvegetable price index, in China. Making Research on the prediction of the index can helpus to grasp the trend of changes in prices of various vegetables, and then makeappropriate adjustments on future behavior and decision-making; therefore, it canpromote the benign development of the vegetable industry.In today’s mathematical theory system, there exist a lot of forecast model has beenused on different kind of research object, however, currently, few scholars has done somequantitative forecast research on Shouguang vegetable price index. Thus, in this thesis,three different kind of forecast model has been used to help separately doing analysis onthe composite index and three level four indexes which selected from the system of theShouguang vegetable monthly given base price indexes, meanwhile, compared theforecast performance of the three different models. In this thesis, the construction ofGM(1,1) model and Bayesian-BP artificial neural network are based on the application ofthe software MATLAB6.5, and the construction of ARMA(or ARIMA) model are basedon the application of the software Eviews6.0, and some figure in this these are drew byR. Through the comparison among the Grey System Forecasting analysis, the time seriesprediction analysis and the Bayesian-BP artificial neural network predictive analysis, wefind out that the ARMA (or ARIMA) model is the best one to get a good prediction effectfor the four price indexes which are selected in this thesis, but its prediction effect are notbetter than the GM(1,1) model in a short term; The more smother the price index is, thebetter the GM(1,1) model is, but the prediction effect is better in a short term than in along term in the GM(1,1) model; the prediction effect of the Bayesian-BP artificial neuralnetwork model is the worst among the three, what may lead to the result is the smallnumber of samples we selected, in detail, the actual data we can collected now is veryless so that it is not enough to make the artificial neural network to be trained and learntwell and extract more useful information to generate accurate forecast map.According to the conclusion of this thesis, we can use the GM (1,1) model forshort-term prediction of some vegetable price indexes, thus to give a reference forcorresponding short-term behavior and decision-making; Meanwhile, we can also use theARMA (or ARIMA) model for long-term prediction of most vegetable price indexes andgive advice to some long-term behavior and decision-making. And there also exist someinadequacies in this thesis, so some further research can be made to forecast the system of the Shouguang vegetable price index more actually on the basis of this thesis, andprovide some advice and guidance for the actual production and living, and promotethe stable development of the vegetable industry.