Abstract: Spatial weight matrix quantificationally expresses the spatial relationships between the geographic features, which is an important tool for spatial statistical analysis. The different methods in generating spatial weight matrix, which is used for spatial autocorrelation analysis in ArcGIS9.3, are analyzed by Hunan's county income gap between urban and rural as a case for empirical analysis. The analysis indicates that with the expansion of average influenced scope of each feature，the global Moran'I shows a curve drop，and the result of local spatial autocorrelation presents significant differences.