Research On A New GDP Forecast Method Based On EEMD-SVR-Copula Model And Nighttime Lighting Data

GDP symbolizes the economic performance and strength of the country.China’s GDP is as large as 82 trillion,and the total volume changes with each passing day.However,accounting methods still have problems such as complex procedures,large statistics,and subjective interferences.Therefore,it is particularly important to obtain efficient,concise and accurate GDP estimation methods.In terms of the saturation phenomenon of the night-time light data,reflecting intuitively the degree of prosperity of the economy,we constructs a de-saturation correction model based on the EEMD-SVR method.In order to prevent the pseudo-regression which is caused by limited sample size and non-stationary data,we presents a new method to construct the probability prediction model of GDP,the EEMD-SVR-Copula model,by using the Copula function and the conditional distribution function.Basing on the model,we forecast the GDP of Shanghai highly developed and Chongqing highly developing,the relative errors between the predicted GDP and the actual values in the two regions are obtained,that is,shanghai is 5.37% and Chongqing is 4.47%.we notice that the prediction accuracy is relatively high.The model has achieved good prediction effect in cities with different degrees of development,indicating the good stability and universality of the model.This paper combines the data decomposition method EEMD,the data synthesis method SVR and the probabilistic prediction method based on the Copula function to construct a new GDP forecasting method.The main advantages of this model have the following three points.Firstly,because the estimation accuracy of the model depends only on the night-time lighting data and model parameters,the influence of subjective factors is decreased in the GDP statistics process;Secondly,compared to other de-saturation correction methods based on the vegetation index NDVI,this model which is based on the data itself without resorting to other variables,is more conducive to the application of nighttime lighting data;Thirdly,compared to other methods that use night-time light data to predict GDP,such as GRNN model,linear regression.this model has less prediction error.In summary,this model provides new idea and way for the prediction of GDP and other socioeconomic parameters,and has important practical value.