Several Methods For Forecasting GDP In Shandong Province And Their Comparison

GDP is an important factor to measure the economic status and development level of a country or region,and the GDP index of the country or region is closely related to inflation,unemployment,and economic growth in the region,and these basic indicators can fully and effectively reflect the future economic development in the region.This will help the government to formulate a macro-control plan in a timely and effective manner based on the forecasted GDP of the region.Therefore,the prediction of GDP has always been the focus of academic research,and it is of great significance to use scientific and effective methods to increase the forecasting accuracy of GDP.As a non-stationary time series,Shandong Province's GDP may contain linear relationships and nonlinear relationships.Therefore,this paper uses four models to judge the relationships contained in the sequence and identify the best model.First,this paper establishes a classic time series analysis algorithm,adopts ARIMA model,determines the optimal parameters of ARIMA model according to AIC criteria,establishes the linear function of GDP in current period and its lag period in Shandong Province,and uses R software to simulate and predict;Secondly,this paper deeply studies the application of BP neural network in time series,selects the appropriate network structure based on the data characteristics of Shandong Province's GDP,selects the nonlinear activation function to use R software for training and learning,and mines the nonlinear relationship of GDP,realizing Shandong Province's GDP forecast;Thirdly,based on the conjecture that the GDP sequence of Shandong Province contains both linear and nonlinear relationships,a combined model based on ARIMA and BP neural network is established in this paper.Based on the linear function of ARIMA model,the nonlinearity of its prediction residuals is excavated.Through simulations and tests,The paper confirms the two characteristics of the sequence.Finally,the combined model is further improved.Based on the linear model established by ARIMA,the BP neural network is constructed by using ARIMA's predictive value and residual sequence to predict the new residual value.The two models are combined to obtain improved combined forecasting model.By comparing and analyzing the forecasting effects of ARIMA model,BP neural network model,combined model,and improved combined model,it is concluded that the optimal model for forecasting GDP of Shandong province is an improved combined model,and uses the model for 2018-2020 Shandong Province GDP forecast.