【摘要】 提出了一种新的时间序列预测方法:利用EMD分解法和神经网络进行信号预测。EMD分解法是一种自适应的信号处理方法,适用于分析非线性、非平稳过程。这一方法分三个基本步骤:首先对待预测的时间序列进行EMD分解,产生多个基本模式分量IMF(Intrinsic Mode Function),接着对各个分量分别用神经网络进行预测,然后重构各分量的预测值,得到最终预测结果。仿真结果表明该方法较之以往的参数模型法以及直接用单一的神经网络进行预测更加准确有效。更多还原

【Abstract】 A new method for time series prediction based on EMD and ANN is presented in this paper.The EMD method is a new method to analyze instability and nonlinearity.It includes three procedures.First,a non-stationary signal is decomposed into some IMFs.Second,predict all the IMFs using ANN,at last reconstruct the signal.The simulation results show this method is more accurate and efficient compared with previous parameter model method and direct ANN prediction.更多还原