介绍分位数回归的原理和方法,相比于传统的沉降预测方法,分位数回归摒除最小二乘回归仅仅分析被解释变量条件期望的缺点,能更好地显示被解释变量的条件分布,适用于隧道沉降监测数据的预测,并结合宁芜货线隧道沉降实例进行拟合和预测,将分位数回归模型预测结果与指数平滑法预测结果进行比较,结果显示分位数回归法预测精度较高。 The principle and method of quantile regression are introduced. Compared with the traditional settlement prediction method, quantile regression excluding least squares regression only analyzes the shortcoming of the condition of the explained variables, and can better display the conditional distribution of the explained variables, Which is suitable for the prediction of tunnel settlement monitoring data and the fitting and prediction of tunnel settlement in Ningwu Expressway. The prediction results of quantile regression model and exponential smoothing method are compared. The results show that the prediction accuracy of quantile regression method Higher.