中文摘要：

回声状态网络（Echo State Network,ESN）能够极好地逼近非线性系统,在非线性混沌时间序列的预测中取得了良好的效果。但是,由于ESN的训练和预测过程是1个高维权值矩阵的运算过程,往往不能获得更好的预测速度。文章提出了一种基于主元分析与回声状态网络相融合的非线性混沌时间序列预测模型。该模型通过主元分析降低输入向量的维数,以减小ESN输入权值矩阵的规模,降低运算的复杂度,从而达到减小ESN训练时间、提高预测速度的目的。利用仿真数据对ESN和文中模型进行了精度和预测时间对比实验,表明该模型是一种有效模型。

英文摘要：

The introduction of the full paper points out that the prediction model based on ESN（echo state network） proposed by H.Jaeger in Ref.1 is,in our opinion,not good enough.So we propose what we believe to be a new and better prediction model based on ESN and PCA（principal component analysis）.Section 1 briefs the relevant information in Refs.2 through 6.Section 2 explains our prediction model;its block diagram is given in Fig.2.Section 3 gives eq.（6） for measuring the prediction precision of our prediction model.Section 4 compares the simulation results of our prediction model with those of the prediction model based on ESN.The simulation results,given in Figs.4 through 7 and in Tables 1,2 and 3,and their analysis show preliminarily that：（1） the computing complexity of our prediction model is lower;（2） its training time is shorter;（3）the prediction rate is higher than that of the prediction model based on ESN.