第9卷第4期 防 灾 科 技 学 院 学 报 Vol.9 No.4 2007年12月 J.of Institute of Disaster-Prevention Science and Technology Dec. 2007 作者简介:王福昌(1974-)男,硕士,防灾科技学院数学教研室讲师,主要从事应用概率统计、优化计算与数学软件研究。 收稿日期:2007-06-26 最小一乘回归系数估计及其 MATLAB 实现 王福昌 胡顺田 张艳芳 (防灾科技学院,河北三河 065201) 摘 要:采用最小二乘准则进行回归分析和数据拟合时,容易受到奇异点的影响,最小一乘准则虽能很好地克服这一缺陷,但又由于它是不可微优化问题,存在计算上的困难。本文对最小一乘准则下线性回归模型分析,转化为线性规划问题,利用 MATLAB 软件快速、准确地估计出最小一乘线性回归的系数,算例说明这种方法是一种有效的参数估计方法。对于非线性问题,可通过线性化而转化为线性问题,用上述方法进行处理。 关键词:最小一乘;线性模型;线性规划;非线性模型;MATLAB 中图分类号:O241.5 文献标识码:A 文章编号:1673-8047(2007) 04-0085-05 The Coefficient Estimation of Least Absolute Deviation Regression and Implementation in MATLAB Wang Fuchang Hu Shuntian Zhang Yanfang (The Institute of Disaster-Prevention Science And Technology, Sanhe, Hebei 065201) Abstract: The least squares regression is vulnerable to outliers, while least absolute deviation (LAD) regression can overcome this shortcoming. However LAD regression coefficients are a problem of non-smooth optimization and difficult in calculation. Linear model is analyzed by translating it to linear programming problem. The linear regression coefficients under LAD criteria are given by solving a linear programming in MATLAB quickly and correctly. The examples show that the method is effective for LAD parameters estimation. The nonlinear regression coefficients under LAD criteria can be dealt with by translating it into linear regression using the method above. Keywords: LAD; linear model; linear programming; nonlinear model; MATLAB 前言 最小二准则在理论与方法
matlab中回归系数,最小一乘回归系数估计及其MATLAB实现
最新推荐文章于 2023-02-07 22:27:19 发布