贝叶斯地理加权回归模型的稳健性研究

Research on Robustness of Bayesian GWR Model

文章在简要介绍地理加权回归(GWR)模型的基础上,推导地理加权回归模型的贝叶斯估计方法(BGWR),并分析该估计方法的稳健性。通过模拟实验研究贝叶斯地理加权回归在数据包含各种异常值情况下的稳健性,结果显示:贝叶斯地理加权回归比地理加权回归具有更好的稳健性。虽然贝叶斯地理加权回归模型计算时间略多一些,但是在空间数据集可能包含异常值的情况下,贝叶斯地理加权回归模型更为可靠和有效。

On the basis of briefly introducing the geographically weighted regression(GWR) model, this paper deduces the Bayesian estimation method of the geographically weighted regression(BGWR) model and analyzes the robustness of the estimation method. And then a simulation experiment is conducted to study the robustness of the Bayesian geo-weighted regression when the data contains various outliers. The results show that the Bayesian geo-weighted regression method is more robust than the geo-weighted regression. Although the Bayesian geographically weighted regression model has slightly longer computational time,it is more reliable and effective when the spatial dataset may contain outliers.