克服复共线性影响的方法探讨

数据建模是参数估计与精度估计的主要手段，也是实际工程中进行系统故障诊断与修复的主要途径。然而，由于变量之间的复共线性性，严重影响了目前广泛应用于工程实际中的最小二乘方法、主成分估计等方法的收敛性与估计精度，也成为制约各种估计方法实用效果的主要原因。本文详细分析了复共线性影响参数估计的机理，全面讨论了克服复共线性影响的几种常用方法的思想及优劣，并给出了工程实际中在参数估计方面应注意的问题。

A Method Discussed to Overcome the Effects of the Compound Linear

The modeling is a main means to estimate parameters and precision, and is also a main approach to diagnose and repair the system's fault in practice. However, the astringency and estimating precision is affected seriously because of the compound linear of variables. This paper discusses the mechanism of compound linear effects particularly and analyzes the subsistent problems of several methods common used in practice, based on this proposed and improved main components analysis approach to overcome the affects of compound linear. The theory and simulation demonstrating the method are utility.