两种成分提取回归方法的比较分析  

Comparing Two Regression Methods Based on Different Principal Components

作　　者：王惠文[1] 杨湘龙[1] 

机构地区：[1]北京航空航天大学

出　　处：《北京航空航天大学学报》1998年第5期575-578,共4页Journal of Beijing University of Aeronautics and Astronautics

基　　金：国家自然科学基金;航空科学基金

摘　　要：对两种采用成分提取进行回归建模的方法进行比较分析．指出采用主成分分析提取的主成分，虽然能很好地概括自变量系统中的信息，却往往对因变量缺乏解释能力．而采用ＰＬＳ回归所提取的成分，则能在很好地概括自变量系统信息的同时，最好地解释因变量，排除无意义的信息干扰作用．This paper deals with the comparing two regression methods based on different principal components.The principal components acquired by principal component analysis are the best summery of the information in the independent variables' system,but generally they are short of the explanatory capability to the dependent variable.However the components acquired by using PLS regression can not only summarize the independent variables well,but also they have the best explanation to the dependent variable.Moreover,the interference from invalid information is eliminated at the same time.

关 键 词：回归分析 比较 主成分分析 PLS1回归 

分 类 号：O212.4[理学—概率论与数理统计]