Comparison Of GDP Forecast Model In Hunan Province

GDP(Gross Domestic Product)is a core indicator in the macroeconomic filed.Its size can reflect the strength of the macroeconomic strength of country or region.When the relevant departments want to adopt proper and effective macroeconomic control,they need to make accurate predictions of the total economic output and trend in the future.According to the forecast results,they judge whether they need to expand or restrain the economic scale at present.In recent years,hunan province has witnessed rapid economic development.In order to formulate effective economic policies in time,it is of great significance to predict GDP.Many domestic scholars have done some research on China's GDP forecast,but they have not compared the prediction effects of multiple models in detail,and there are few studies on the GDP forecast of hunan province.In this paper,a variety of methods are adopted to establish the GDP forecast model of hunan province,and the prediction effects of each model are compared and analyzed in detail.It includes three common models in traditional time series analysis,ARIMA,exponential smoothing,trend curve models,and BP neural network models in machine learning methods.Among them,when using the BP neural network to predict Hunan's GDP,two research ideas are used.One is to model GDP as a time series,and the other is to use the indicator variables that affect GDP as input variables to predict the value of GDP.There are many factors that affect GDP.In this paper,we use the Adaptive-Lasso method to screen and compress many factors that affect GDP..In addition,the ARIMA model and BP neural network model are combined for combined prediction in a certain way.Finally,the optimal model is selected to predict the GDP of hunan province from 2019 to 2021.There are some conclusion based on the empirical analysis of Hunan GDP:1 ? The stability of quadratic exponential smoothing prediction model and the cubic curve prediction model is poor,and with the increasing of forecast period,the forecast error is increasing,so it is not suitable for long-term prediction of GDP,but there is a certain reference value for the short-term prediction.2?Both BP neural network models have a good prediction effect and stable prediction performance,which is suitable for long-term GDP prediction.3 ? the combination model improves the prediction ability of single model significantly,with higher accuracy and better stability.The predication ability of the six model is sorted in descending order as:combined prediction model,BP neural network model based on index variable,BP neural network model based on time series,secondary exponential smoothing model,ARIMA model,curve model.