ARIMA与BP神经网络混合模型在广西GDP预测中的应用

为改进单纯使用ARIMA模型或BP神经网络模型对GDP预测的效果,笔者以1978—2008年的广西GDP为样本,首先建立ARIMA模型,得到拟合误差序列及2009—2015年的广西GDP的初始预测值,再对误差序列构建BP神经网络并得到2009—2015年的误差预测值,最后,用误差预测值对初始预测值进行修正,得到修正后的2009—2015年广西GDP的预测值。结果表明,ARIMA与BP神经网络混合模型的预测结果显著优于单一模型的预测。

The Application of ARIMA and BP Neural Network Mixture Model in the GDP Estimation of Guangxi

In order to improve the effect of predicting GDP simply by using ARIMA model or BP neural network model,this essay uses the GDP of Guangxi from 1978 to 2008 as the sample,firstly establishes the ARIMA model to get the fitting error sequence and the initial forecast of the Guangxi GDP from 2009 to 2015,and then builds up the BP neural network for the error sequence to get the error predictive value from 2009 to 2015,finally uses the error predictive value to revise the initial predictive value to get the revised predictive value.The result shows that the application of ARIMA and BP Neural Network Mixture Model in the estimation of Guangxi＇s GDP is significantly better than the forecasts by single model.