根据地铁隧道监测点沉降变化中非线性、不确定、时变性的特点,建立了基于小波分析的支持向量机预测模型。首先运用小波分析将监测点沉降序列分解为低频近似分量和高频细节分量,然后对各分量分别进行支持向量机预测,最后将各分量预测结果进行小波重构得到监测点的沉降预测曲线。预测结果表明,在相同样本数和短周期预测条件下,Wavelet-SVM模型的预测精度优于BP神经网络方法。对地铁沉降监测提前进行预警预报有一定的参考价值。 According to the non-linear, uncertain and time-varying characteristics of subsidence changes of subway tunnel monitoring points, a support vector machine prediction model based on wavelet analysis is established. Firstly, the settlement of monitoring points is decomposed into low-frequency approximation components and high-frequency details components by using wavelet analysis. Then, SVM is used to predict each component separately. Finally, the prediction results of each component are reconstructed by wavelet to get the settlement prediction curve of monitoring points. The prediction results show that the prediction accuracy of Wavelet-SVM model is better than that of BP neural network under the same number of samples and short-period prediction. It is of certain reference value for early warning and forecasting of subway settlement monitoring.