最小二乘回归支持向量机对非线性时间序列预测的试验分析

利用最小二乘回归支持向量机LS-SVMR(least square support vectors machines for regression)对2个不同长度的时间序列资料,国家气候中心1982年1月～2005年12月Nino3区逐月海温距平指数(短序列),及1950年1月～2006年12月Nino3区逐月海温距平指数(长序列)资料进行了预测试验,以验证支持向量机对气候变化中非线性时间序列的预测效果.结果表明:通过训练建立的最小二乘回归支持向量机模型,较好地反映了Nino3区海温距平指数的变化规律,36个月的预报效果较好,具有一定的可信度.资料的长度越长,预测结果与实测值的变化趋势越接近,但资料长度对均方根预报误差不敏感.

Experiments on nonlinear time series prediction with least square support vector regression machine

In order to verify the prediction results of support vectors machines(SVM) for the nonlinear time series data in climate changes,the time series data of two different lengths from National Climate Center,which were ASST indices in Nino3 areas from January 1982 to December 2005,and from January 1950 to December 2006 were used to predict their future tendency by utilizing Least Square Support Vector Regression Machine(LS-SVRM).The results show that LS-SVRM models established by training historical data can ef...