Prediction Of China's Inflation

The formation mechanism of inflation is more complex and the factors affecting inflation are more diverse.How to grasp the economic situation and transmission mechanism to make accurate and efficient prediction of inflation in China is a difficult problem facing the government.The existing research on inflation prediction are mostly based on data with the same frequency.When comes to the data with different observation frequencies,they need to convert it to same frequency in advance to be included in the model.This method affects the prediction accuracy of the model and the empirical results are not convincing.As a new non-parametric statistical method,the functional data analysis method can solve the problem of model estimation when the time interval between variables is inconsistent and data is missing,which is difficult to deal with by traditional models.It can be used for more relaxed assumptions and is not restricted by parameters.It has strong applicability and simple calculation,which makes it convenient for us to introduce high-frequency financial data variables into the model for inflation prediction.Under the premise of making more effective use of existing information and taking into account the use of derivative information,the prediction accuracy of inflation is expected to be further improved under the guidance of this method.From the perspective of functional data analysis,this paper research on the dynamic change behavior and prediction problem of inflation in China.Using functional regression analysis method,this paper constructs a functional linear regression model to predict the provincial CPI of 31 provinces in China with mixed daily,monthly and quarterly data.Subsequently,the first and second derivatives of the provincial CPI curve smoothed by the basis function method are obtained,which are introduced into the original model as speed and acceleration factors respectively,then the functional linear regression model with dynamic factors is constructed for comparison and prediction.In order to verify the validity of the model,the final prediction results of the functional linear regression model are compared with those of the functional linear regression model with dynamic factors,the functional time series model and the traditional time series ARIMA model by using the indexes of root mean square error and average percentage error.The results show that the functional linear regression model constructed in this paper can predict inflation with high accuracy,and its prediction effect is better than that of functional linear regression model with dynamic factors,functional time series model and ARIMA model.In addition,the prediction effect of functional time series model is better than that of traditional time series ARIMA model,which confirms that the prediction accuracy of the traditional time series model has been improved to a certain extent through this method.However,the prediction error of the functional linear regression model with dynamic factors is relatively large,and the introduction of derivative information can not meet the expectation of improving the prediction accuracy.The model still needs further improvement in the follow-up study.