线性EIV模型参数的LASSO估计方法  

The LASSO estimation method for linear EIV model parameters

作　　者：赵明清 席甜甜 ZHAO Mingqing;XI Tiantian(College of Mathematics and Systems Science,Shandong University of Science and Technology,Qingdao 266590,China)

机构地区：[1]山东科技大学数学与系统科学学院,山东青岛266590

出　　处：《山东理工大学学报（自然科学版）》2022年第3期69-73,共5页Journal of Shandong University of Technology:Natural Science Edition

基　　金：山东科技大学研究生导师指导能力提升计划项目(KDYC17018)。

摘　　要：基于结构风险最小化原则,提出了可以实现高维数据降维的线性EIV模型参数的LASSO估计(LE)方法,并给出了其数值解的迭代算法。为说明LE方法的有效性,通过实证与WTLS、LS两种方法进行了对比分析。结果表明,LE方法能够明显提高预测精度,具有更强的泛化能力,同时可以实现变量选择,达到高维数据降维的目的。In this paper,considering the principle of structural risk minimization,the LASSO estimation method for linear EIV model parameters(LE),which can achieve dimensionality reduction,is proposed,and its numerical solution with iterative algorithm is given.In order to illustrate the effectiveness of the LE method,it was campared with WTLS and LS methods.The results showed that the LE method proposed in this paper can significantly improve the prediction accuracy,and achieve high generalization and variable selection to fulfill the purpose of dimensionality reduction of high-dimensional data.

关 键 词：EIV模型 参数估计 结构风险最小化原则 LASSO 

分 类 号：O212[理学—概率论与数理统计]