主成分分析法在神经网络集成预报中的应用

APPLICATION OF PRINCIPAL COMPONENT ANALYSIS METHOD IN NEURAL NETWORK INTEGRATIVE FORECASTING

本文将主成分分析方法引入到用神经网络作集成预报中,根据各个主成分的方差贡献的大小来确定所需主成分的个数,在此基础上,对集成预报的原始因子进行重构,以达到对数据进行降维,加快计算速度和提高预报准确率的目的。

This paper applies Principal Component analysis method to neural network integrative forecasting, and decides the required number of PC according to contribution of each PC to variance, then reconstructs the predictand to accelerate speed of calculation and improve the accuracy of prediction.