一种选权迭代法的半参数回归模型在滑坡预测中的应用

The Application of Selecting Weight Iteration for Semiparametric Regression Model in Landslide Prediction

传统抗差估计中的选权迭代法无法探测到模型的系统误差,而基于补偿最小二乘原理的半参数模型可以较好地分离出系统误差。通过建立选权迭代法的半参数回归模型,利用时间序列法、L曲线法分别确定模型中的正则化矩阵及平滑因子,并利用选权迭代法重新定权,同时降低了观测粗差和系统误差对参数估值的影响。通过仿真算例,并以重庆奉节县大坪滑坡实测数据为例,验证了选权迭代法的半参数回归方法应用到三峡库区滑坡预测的有效性和优越性。

In traditional robust estimation,the selecting weight iteration method can not detect the systematic errors,while the semiparametric model based on the least square principle can separate the systematic errors well.This paper establishs the semiparametric regression model based the selecting weight iteration method,using the time series method and L-curve methods to determine the regularization matrix and smoothing factor,and uses the selecting weight iteration method to reshape the weight matrix so as to reduce the influence of gross errors and systematic errors in parameter estimation.Using a simulated example,and taking the measured data of Daping landslide in Fengjie county of Chongqing as an additional example,the validity and superiority of the selecting weight iteration for semiparametric regression method applied to landslide prediction in the Three Gorges reservoir area is verified.