Prediction Of CPI Index Based On SARIMAX-EEMD-LSTM Model

Since the time series of consumer price index(CPI)in Shandong province has both nonlinear and linear characteristics,predicting its index is not easy,and the traditional forecasting methods cannot accurately predict the consumer price index.In order to make accurate forecasts of CPI,this paper takes the production price index(PPI)as an exogenous variable and uses a seasonal differential autoregressive moving average(SARIMAX)model with exogenous variables.And the SARIMAX model is mixed with the long and short term memory network(LSTM)model based on integrated empirical mode decomposition(EEMD)under which the SARIAMX-EEMD-LSTM hybrid model is proposed for CPI forecasting.An empirical study is conducted with the consumer price index in Shandong Province to verify that the proposed forecasting framework is feasible.The framework shows better accuracy compared with the forecasting results derived from other forecasting models.The hybrid model can provide a more practical and reliable theoretical support for the proposed CPI-related policies.In this paper,in order to uniformly test the strengths and weaknesses of different models for CPI forecasting,monthly year-on-year CPI data for Shandong Province from September 2000 to January 2022 are applied from the beginning to the end.First,this paper constructs and forecasts the data using a single SARIMA model,SARIMAX model,and LSTM model,respectively.With the help of Granger causality test,this paper finds that the SARIMAX model with PPI as an exogenous variable has the best prediction effect.Then,the residuals generated from the original CPI series and the predicted series from the SARIMAX model are decomposed to obtain several eigenmodular functions(IMFs)and a residual series with trend terms.After that,the LSTM model is applied to each sequence separately to fit the prediction,and the prediction results of each sequence obtained are summed to obtain the prediction results of the whole residuals,and finally combined with the prediction results of the SARIMAX model,the final CPI prediction results are obtained in this paper.At the end of the paper,by comparing the numerical magnitude of the errors and analyzing the advantages and disadvantages of the prediction results of each model,the best hybrid model of CPI prediction under other models is obtained,which is called SARIMAX-EEMD-LSTM hybrid model.Finally,by analyzing the model of SARIMAX-EEMD-LSTM,a hybrid model of "dec-omposition and prediction"is proposed based on the linear and nonlinear models,and its prediction accuracy is significantly improved.This finding also proves in a sense the application of SARIMAX-EEMD-LSTM hybrid model in CPI analysis and forecasting.It provides a theoretical basis for the government of Shandong Province to understand the current economic situation in Shandong Province and to formulate corresponding economic policies.