基于空间相关的逻辑回归模型的城市扩展模拟

Simulation of Urban Expansion Based on Logistic Regression with the Consideration of Spatial Correlation

建立一种基于空间的逻辑回归模型用于分析和预测城市扩展,这种模型利用空间滤波对因子进行空间平滑得出"隐含变量",从而把原始变量变换为没有空间相关的新变量.利用该模型对新堡市的土地利用变化进行了分析和预测.实验结果表明,新方法比普通逻辑回归模型的预测精度有较大提高,曲线下面积从0.74提高到0.84.

This paper proposes a space-based logistic regression model to analyze and predict urban expansion. This model uses the spatial filtering of the spatial smoothing factor to produce the ＂hidden variables＂ ,thus transforms the original variables to new variables without spatial correlation. The proposed model is employed to analyze and predict urban expansion. The results indicate that the accuracy of proposed method is better. The area of AUC rises from 0.74 to 0.84.