提出了一种新的时间序列组合预测模型(PCA-BPNN模型),即先利用主成分分析方法对原输入变量的空间进行重构,然后借助于各主成分对总体样本的贡献率来确定网络结构.最后利用1999年1月至2005年12月甘肃省天水市胆结石发病率的资料验证了该方法的有效性.
A new kind of time series combined prediction model (PCA-BPNN model) is proposed, which the principal component analysis is used to reconstruct the original input space, and the network structure is deter- mined based on the contribution rate of all main components. Finally, the incidence rate of gall-stone from Janu- ary 1999 to December 2005 of Tianshui, Gansu province is used to validate the method.