回归模型中哑变量的相对重要性指数

发布时间：2018-03-17 08:18

本文选题：定性属性　切入点：回归方程　出处：《计算机应用》2017年11期 　论文类型：期刊论文

【摘要】：为在回归模型中描述定性属性,通常需要引入哑变量。对含哑变量的回归方程,提出描述不同哑变量在回归方程中不同重要程度的方法。该方法分解出含哑变量的回归方程中哑变量部分和非哑变量部分的回归平方和,计算这两部分在该回归方程中所起作用的占比,将该占比设计为各哑变量在回归方程中的相对重要程度指数。在近10万笔的Lending Club和Prosper网络借贷数据集上,所进行的挖掘借款用途对借款成功率、信用等级对借款利率的影响程度的实验结果表明,与传统回归方程仅提供哑变量前的系数却不能展现其重要程度相比,所提方法展现出不同哑变量的不同重要程度,为定量分析回归方程中定性自变量对因变量的影响程度提供了重要的手段。
[Abstract]:In order to describe qualitative attributes in regression models, it is usually necessary to introduce dummy variables. This paper presents a method to describe the different importance of different dummy variables in the regression equation, which decomposes the regression square sum of the dummy variable part and the non-dummy variable part in the regression equation with dummy variable. The ratio of these two parts in the regression equation is calculated, and the ratio is designed as the relative importance index of each mute variable in the regression equation. On nearly 100,000 Lending Club and Prosper network loan data sets, The experimental results on the influence of loan usage on loan success rate and credit grade on loan interest rate show that compared with the traditional regression equation, the coefficient before the dummy variable can not show its importance. The proposed method shows different degrees of importance of different dummy variables and provides an important means for quantitative analysis of the influence of qualitative independent variables on dependent variables in regression equations.
【作者单位】：福建师范大学数学与信息学院;福建省网络安全与密码技术重点实验室(福建师范大学);
【基金】：国家自然科学基金资助项目(61672157) 福建师范大学网络与信息安全关键理论和技术创新团队项目(IRTL1207)~~
【分类号】：O212.1

【相似文献】