论文部分内容阅读
主成分分析在一定程度上可以解决大坝变形监测回归模型因子间的复共线性,然而当提取的主成分信息不充分时,主成分回归用于大坝安全预测可能失效。提出以主成分分析提取的主成分作为半参数回归的参数分量,剩余成分和模型误差作为未知的非参数分量对主成分回归进行补偿,建立一种基于主成分和半参数的大坝变形监测混合回归模型。利用某大坝实测资料进行建模分析,结果表明该混合模型能克服回归因子间的复共线性,避免半参数回归补偿最小二乘估计中法矩阵的病态性,比传统的主成分回归和逐步回归模型具有更好的拟合和预报精度。
The principal component analysis can solve the complex collinearity between dam deformation monitoring regression model factors to a certain extent. However, when the principal component information extracted is not sufficient, the principal component regression can not be used in dam safety prediction. The main component extracted by principal component analysis is proposed as the parameter component of semi-parametric regression, the residual component and model error as the unknown non-parametric component to compensate the principal component regression, and a dam deformation monitoring hybrid based on principal component and semi-parametric Regression model. The results of modeling analysis based on measured data from a dam show that the hybrid model can overcome the complex collinearity between regression coefficients and avoid the ill-posedness of the matrix of law in the least squares estimation by semi-parametric regression. Compared with the traditional principal component regression and stepwise Regression model has better fitting and forecasting accuracy.