基于PLS成分的变量筛选法

The Variable Selection Based on PLS Component

摘要:

就多元线性回归模型，提出一种基于PLS成分的变量筛选法。其基本思想是利用偏最小二乘回归原理，提取系统的两个主要成分，然后根据变量入选原则，保留对因变量影响较大的变量；再利用PLS成分筛选剩下的变量，经有限次筛选，即可得到最佳变量子集。应用算例证实了这一方法的有效性和可行性。

Abstract:

A variable selection method based on PLS Component is proposed for multiple linear regression model. The two dominating components of the multivariate system are extracted according to the partial least-square regression principle, then the variable which