Empirical Analysis Of Chongqing GDP Based On ARIMA And BP Neural Network Model

As an important indicator for measuring the economic development of a country and a region,GDP is quantitatively analyzed with relevant knowledge to understand its characteristics and to grasp its development trend.Relevant departments can formulate relevant economic policies to regulate and control according to this value.By analyzing the GDP data,it is also possible to test the scientific nature and effectiveness of the formulated macro-policies,and it is practical to accurately grasp the future economic development trends.This paper introduces the theory and practical application of ARIMA model,BP neural network and combination model.It mainly finds the optimal model of fitting GDP by empirical analysis of Chongqing GDP,and then makes prediction based on the optimal model.The GDP of Chongqing from 1997 to 2017 was selected as the research sample.The GDP data of 1997-2014 was used as the training sample,and the GDP data of 2015-2017 was used as the test sample.The average absolute percentage error(MAPE)and the Tyre coefficient were used.Four standard evaluation criteria of prediction error(SDE)and prediction accuracy are used to evaluate the model test results.Firstly,the optimal parameters of the model are determined according to the AIC criterion,and the optimal model ARIMA is determined.Secondly,when constructing BP neural network,the momentum factor is added to optimize its network.According to the test results,the neural network fitting error is used again to construct a neural network to further correct the error to fully extract the sequence information.Finally,using the equal weight method,simple weighting method,error squared and reciprocal method,and mean squared error variance,the weights of the individual models are given to form a new combined model,and then the seven models are compared and analyzed to obtain the best model.The empirical analysis shows that both the ARIMA model and the modified BP neural network are suitable for fitting Chongqing GDP.The modified neural network is better than the ARIMA model and the per-corrected neural network model.From the analysis of the combined model results,it can be seen that the combined model can improve the accuracy of the single-item model to a certain extent,which is better than the single model.When a single model is combined,the weights are given differently and the results are different.Finally,through the comparative analysis of the seven models involved in this paper,the optimal model is the modelthat combines the ARIMA model with the modified neural network by the combination of the error variance mean square reciprocal method.Finally,the final model was used to forecast the GDP of Chongqing city in 2018-2020.