动态回归模型在变形分析中的应用

摘要

The recursive least squares algorithm for dynamic regression model, which can be used in the online regression modeling for dynamic data sets, is deduced. When the new observations updating, the dynamic regression algorithm, by avoiding the increment of matrix size, and avoiding the computation of the inverter matrix, can decrease the computation time of solving model parameters. The algorithm is easy to be programmed and no iterative computation is needed. The simplicity and practicality have been verified by using the dynamic regression approach in Zhexi and Dongjiang Dam deformation prediction samples. By compared with other algorithms, the method can reduce computation time and improve prediction precision.%为了使回归模型适应动态数据集,推导了动态回归模型的递推最小二乘算法,数据更新时,采用修正方式更新回归系数计算两个矩阵,避免了重复矩阵求逆运算,实现了观测数据增加而矩阵阶数不增加,理论上减少了计算时间.以柘溪和东江两大坝变形分析多元动态回归模型为例进行实验,结果表明:该方法建模过程简洁、无需迭代计算,易于编程实现,在计算效率与预报精度等方面均具优势,可应用于实时变形分析建模.