混合地理加权回归模型算法研究

以迭代算法为基础，推导出混合地理加权回归模型的常系数（全局参数）和变系数（局域参数）的计算方法，并以上海市住宅小区楼盘销售平均价格为例进行验证。结果表明，混合地理加权回归模型的计算量略大于地理加权回归模型，但对样本数据的拟合更好，局域参数估计更稳健。

Algorithm for Mixed Geographically Weighted Regression Model

An iterative algorithm is developed to estimate global coefficients and local coeffi-cients in MGWR.First independent variables are classified two groups,Groupagin which variables are global associated with global coefficients and Groupbgin which variables are lo-cal associated with local coefficients.Second assuming thatagis known,coefficients ofbgiscalibrated by using the basic GWR.Third ordinary linear regression(OLR) is used to esti-mated coefficients ofag.Material formulations of two types of coefficients and computational progress are also produced,and further tested by using average prices of house blocks in Shanghai.The experiment proves that all formulations of coefficients are available,and com-parison of the two models by Akaike information criteria value shows MGWR is more appro-priate and stable for the local coefficients estimates than BGWR although it requires a greater computational effort.