摘要
在回归分析中,自变量之间的多重共线性会严重影响回归效果,因此消除多重共线性成为回归分析中参数估计的一个重要环节。该文探讨了消除多重共线性现象的偏最小二乘回归法,并基于VB语言实现了程序的设计,最后结合某大桥GPS控制网进行高程异常拟合计算,通过与专业软件计算结果的对比,验证了该程序的可行性。
In regression analysis,the multicollinearity between the independent variables will seriously affect the regression results,thus multicollinearity elimination has become an important part in parameter estimation in regression analysis. In this paper,the partial least squares regression method which can eliminate multicollinearity phenomenon is discussed,and the corresponding program is implemented,which is based on VB（ Visual Basic）. Finally,combined with a bridge＇s GPS control network,the abnormal elevation fitting is calculated. The feasibility of the program is verified by comparison with the result which is calculated by professional software.
出处
《勘察科学技术》
2015年第1期47-51,共5页
Site Investigation Science and Technology
关键词
偏最小二乘回归
多重共线性
交叉有效性检验
简化算法
二次多项式曲面模型
高程拟合
partial least squares regression
multicollinearity
cross validation（CV） test
simplified algorithm
quadratic polynomial surface model
elevation fitting
