基于小波分析AR（P）-SVR组合模型在大坝变形预测中的应用

摘要

Low frequency and high frequency components are obtained through wavelet decomposition .The low frequency components are adopted in the AR(P) model to make predictions ,while the high frequency components make predictions with SVR model . Then the predicted data after reconstructing them are obtained .Results show this model has higher prediction accuracy than SVR model without wavelet decomposition and the model that only uses SVR to predict each component after wavelet decomposition .%监测序列经小波分解后，得到低频分量和高频分量。对低频分量采用自回归AR（P）模型预测，对高频分量采用支持向量回归机SVR模型预测，最后将各分量进行小波重构，得到监测序列的预测值。结果表明，此种预测方法比直接使用SVR模型或经小波分解后再采用SVR模型预测精度高。