EEMD、CEEMD算法与SVM在SST时间序列研究中的应用

海洋表面温度(SST)具有非线性、非平稳等特征,给处理和预测带来了很大的困难.将集合经验模态分解(EEMD)、改进的集合经验模态分解(CEEMD)与支持向量机(SVM)方法相结合,实现了对东北太平洋月平均海温距平序列(SSTA)的预测:首先应用EEMD或CEEMD方法将SST数据分解为多个本征模态函数(IMFs),然后应用SVM算法对各IMFs进行拟合、预测,最后对各IMFs预测结果叠加重构得到预测结果.EEMD-SVM和CEEMD-SVM数值模拟结果显示,预测最大误差小于0.25℃,并且CEEMD-SVM预测效果更好,为SST实际预测提供了参考.

The Application of Ensemble Empirical Mode,Complementary Ensemble Empirical Mode and SVM in SST Time Series Research

The sea surface temperature(SST) is nonlinear and non-stationary,which brings great difficulties to the treatment and prediction.In this paper,the combination of Empirical Mode Decomposition (EEMD),Complementary ensemble EMD (CEEMD) and Support Vector Machine (SVM) is applied to predict the SSTA in the Northeast Pacific Ocean.Firstly,the EEMD or CEEMD method is used to decompose the SST data into a number of intrinsic mode functions (IMFs) Then,the IMFs are fitted and predicted using the SVM.Finally,the prediction results are superimposed on the IMFs.And we analyse them.The results of EEMD-SVM and CEEMD-SVM show that the predicted maximum error is less than 0.25℃,and CEEMD-SVM has better prediction effect,which provides a reference for the actual prediction of SST.