Application Of Combined Forecasting Model In CPI Forecasting

Consumer Price Index(CPI),abbreviated as CPI,is an important indicator of statistics,mainly reflects the consumer purchase of goods or services,the price level and price changes in the trend or range.CPI level directly reflects the level of inflation in a country,CPI also involves economic development,is the national managers to develop macroeconomic policies,planning the central bank open market business and analysis of the currency and bond market an important reference.Thus,the study of CPI has important practical significance.In this paper,we use the individual model(SARIMA model,GM(1,1)model and BP neural network model)and the combined model to forecast the monthly CPI of January 2001-January 2017.Combined forecasting model Firstly,the CPI sequence is decomposed into the trend variable and the period fluctuation term by the median moving average method.Then,the forecast term of the SARIMA model is constructed and the GM(1,1)model is constructed.Then,the product model And then the BP neural network model is used to predict the random perturbation term.Finally,the prediction values obtained by the three models are combined by the first product addition method,and the final prediction value is obtained.By comparing the prediction effect of each individual model and the combined model,the optimal CPI forecasting model is found out.The results show that the combination model based on SARIMA-GM-BP has a great improvement in the prediction ability and has a certain ability to resist the influence of random interference.It is proved from the empirical point of view that the combination model can be used in the CPI The prediction effect of the model is not good,that is,each forecasting model has its own scope of application,according to the actual situation application;based on the combined model extrapolation prediction results show that the CPI was moderate The trend of decline.