测绘科学技术学报2017，Vol.34Issue(6)：654-658,5.DOI:10.3969/j.issn.1673-6338.2017.06.021
一种基于主成分分析的时空地理加权回归方法
A Temporal-Spatial Geographic Weighted Regression Method Based on Principal Component Analysis
摘要
Abstract
Aiming at the problem that the spatial-temporal geographical weighted regression method has higher computational complexity when the input variables are more and the prediction accuracy is lower when the input variables are less,a spatial-temporal geographical weighted regression method based on principal components analysis (PCA-GTWR) is proposed in this paper.This method uses the nonlinear principal components analysis method to reduce the dimension of several related variables which affect the PM2.5 concentration,and obtains several comprehensive indexes as the input variables of the GTWR model to predict.In order to verify the effectiveness of the method,PM2.5 data from April 2014 to March 2017 in Beijing is adopted.The Pearson correlation coefficient method is used to select the influencing factors which have higher correlations with PM2.5 concentration as input variables of the conventional GTWR model,and the method of this article is compared in the premise of the same number of variables.The results indicate that through the nonlinear principal components analysis method to preprocess the related variables,it can effectively solve the collinearity between variables,retain the major information of the original influencing factors and improve the efficiency of the algorithm.Besides,the MAE,RMSE,AIC of this method which are lower than those of the conventional GTWR model.In this paper,the goodness of fit reached up to 88.11％.关键词
主成分分析/地理加权回归模型/时空地理加权回归模型/细颗粒物/Pearson相关系数Key words
principal component analysis/GWR/GTWR/PM2.5/Pearson correlation coefficient分类
天文与地球科学引用本文复制引用
卢月明,王亮,仇阿根,张用川,赵阳阳..一种基于主成分分析的时空地理加权回归方法[J].测绘科学技术学报,2017,34(6):654-658,5.基金项目
基础测绘支撑项目(2016KJ0104) （2016KJ0104）
中国测绘科学研究院基本科研业务费项目(7771614). （7771614）
