Because the time series of the GDP has both linear and nonlinear characteristics, traditional forecasting methods, such as neural network and some integrated model, tend to bring errors. BP neural network model using the annual incremental rate of GDP for the network prediction, was set up to predict the GDP. The GDP forecastting results from improved BP model was compared with ARIMA-BP and single BP model, showing more accuracy. '%GDP时间序列具有线性和非线性的双重特征,所以传统统计预测方法、神经网络方法和集成预测方法都在预测分析时准确性不高,误差较大.文章提出由GDP时间序列,找出只具有非线性特征的GDP年增量百分比序列,以此建立基于BP的预测模型,对我国的GDP进行预测,仿真实验表明,改进的BP模型预测准确率明显优于目前的ARIMA-BP集成模型及单一BP模型的预测准确率,从而证实了改进的BP模型用于GDP预测的有效性.
展开▼