基于典型相关的线性回归模型  

Linear regression model based on canonical correlation

作　　者：赖艺芬[1] 梁飞豹[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350002

出　　处：《福州大学学报（自然科学版）》2004年第4期438-441,共4页Journal of Fuzhou University(Natural Science Edition)

摘　　要：通过实验、分析、并理论证明,在满秩情况下,由典型相关求解线性回归模型与直接求解线性回归模型的结果是一致的.将该方法推广到不满秩的情况,提出了一种新的求解线性回归模型的方法,虽求解结果牺牲部分的拟合度,但在一定程度上增强了预测效果.In the case of full rank, we prove the consistent result between the method that solves the linear regression model via canonical correlation and the way which directly solves it. Furthermore, the paper popularizes the method for the situation of non-full rank and proposes a kind of new method to solve linear regression model. The method sacrifices partial fitness to strengthen the efficiency of forecast.

关 键 词：线性回归 典型相关 预测 理想状态 变量 

分 类 号：O21[理学—概率论与数理统计]