以乌尔逊河为例,采用相空间重构理论计算实测月降雨的延迟时间、嵌入维数、G-P饱和关联维数和Laypunov指数,证明月降雨时间序列存在混沌现象。运用LS-SVM模型对乌尔逊河月降雨混沌时间序列进行预测,并利用交叉验证法求取LS-SVM模型两个重要参数的最佳组合,同时与RBF神经网络模型进行了对比分析。结果表明,在做混沌时间序列分析时LS-SVM模型的预测精度优于RBF神经网络模型。 Taking Wuerxun River as an example, phase delay reconstruction time, embedding dimension, G-P saturation correlation dimension and Laypunov exponent were calculated by phase space reconstruction theory to prove the chaos phenomenon of monthly rainfall series. The LS-SVM model was used to predict the monthly rainfall-chaotic time series of the Wuerxun River. The best combination of two important parameters of the LS-SVM model was obtained by cross-validation, and compared with the RBF neural network model. The results show that LS-SVM model has better prediction accuracy than RBF neural network when doing chaotic time series analysis.