An Empirical Analysis Of CPI With ARIMA And BP Model

The consumer price index is macroeconomic index,which reflect price changes of goods and services they buy. Its rate of change in a certain extent, not only reflects the degree of inflation or deflation, but also is a index which the market main body of economic activities and government formulate important reference index of macroeconomic policy. The CPI stability, employment and GDP growth is often the most important social and economic goals. So researching the CPI is meaningful and actual.This paper not only deeply analysis prediction of characteristics and advantages of the ARIMA model and Back Propagation model, but also establish time series prediction model which integrating the ARIMA model with the BP model. This idea is taken advantage of the two kinds of model in linear space and nonlinear space. First of all, The consumer price index time series is decomposed into two parts, this two parts is the linear and the nonlinear residual autocorrelation subject.Next, using ARIMA model predict the linear component of CPI time series, and using the BP model to estimate the nonlinear residuals component. The final result shows that the integrated model prediction is more accuracy than single prediction of BP model, which approved the effectiveness of the integrated model is better which integrate the ARIMA model and BP model to predict the CPI.This paper analyzes the factors which affect the CPI, based on the cointegration theory, cointegrating test for CPI and its influence factors, determine the CPI, GDP, residents’ disposable income, investment in fixed assets are first-order single whole. Because have a cointegration relationship between the sequence, so this paper establishing cointegration and error correction model,forecasting the CPI and compared with the real value. The end, confirming short-term prediction for the CPI, error correction model is higher accuracy than the cointegration model.