基于主成分分析和贝叶斯正则化BP神经网络的GDP预测

选用财政收入、财政支出、消费品零售总额、实际利用外资、进出口总额以及全社会固定资产投资等对GDP有显著影响的6个因子,用1985～2008年中国的宏观经济数据建立了一个基于主成分分析和贝叶斯正则化BP神经网络的预测模型,并把它应用于我国GDP的预测。实证结果表明:通过主成分分析法和贝叶斯正则化方法对BP神经网络进行改进,可简化网络结构,增强泛化能力。与其它常用的预测方法相比,该方法数据输入简便,收敛速度快,拟合曲线光滑,且在预测精度上有明显的优势。

GDP Prediction Based on Principal Component Analysis and Bayesian Regularization BP Neural Network

We choose financial income,financial expenditure,total retail sales of consumer goods,actually used foreign investment,total import and export volume and social fixed assets investment,such as six factors,which have a significant effect on GDP. A forecasting model based on principal component analysis and Bayesian regularization BP neural network was established by using the Chinese macro-economic data in 1985~2008, and was applied to predict the GDP of China. The empirical results show that the principal component analysis and Bayesian regularization are utilized modify BP neural network, which can simplify network structure and strengthen generalization. Compared with other commonly used methods of forecasting, this method has simple data input,fast convergence rate,smooth fitting curve,and there is significant advantage in the prediction accuracy.