摘要
以全国31个主要城市的空气污染数据作为研究对象,通过B-样条逼近离散数据,拟合日变动曲线作为函数型数据,采用基于Group LASSO的函数型变量选择方法对空气污染因素和气象因素进行变量选择,选取影响空气质量的主要因素,并建立函数型回归模型,同时与函数型主成分回归方法进行比较,发现Group LASSO函数型变量选择方法不仅对群组变量进行选择,而且对选择出的群组内变量还能做进一步的选择,可以获得较为稳健的估计结果和较好的选择效果。
In this paper, we use the B-spline to approximate the discrete data, and then fit out the daily fluctuation curve as the ＂data＂ of the model analysis. On this basis, we establish the function variable regression model. On the basis of the functional regression model, the variables of air pollution and meteorological factors are chosen as function variables, and the air pollution and meteorological factors are selected as the function variables, especially as the main factors of affecting air quality. The function variable selection method does not only select the group variable, but also makes a further choice for the selected group variable. The method can obtain more robust estimation results and better selection results.
出处
《安庆师范大学学报（自然科学版）》
2017年第4期7-12,共6页
Journal of Anqing Normal University(Natural Science Edition)
基金
浙江省自然科学基金(LY14A010022)
