提出一种新的应用支持向量机回归原理的混沌时间序列非线性预测方法,同时利用自适应的方法对支持向量机的参数进行优化.仿真结果显示支持向量机具有比传统的回归方法更好的泛化能力,预测方法具有很高的预测精度,同时还讨论了支持向量机中参数以及嵌入维数的变化对泛化误差的影响,得出的结论与统计学习理论中的VC维理论相一致. A new chaotic time series nonlinear prediction method based on the principle of support vector machine regression is proposed, and the parameters of the SVM are optimized by using adaptive method. The simulation results show that the SVM has better performance than the traditional regression method The generalization ability and prediction method have high prediction accuracy. At the same time, the influence of the variation of parameters and embedding dimension in SVM on the generalization error is discussed. The conclusion is consistent with the VC dimension theory in statistical learning theory .