基于多尺度空间聚类的经济区域划分研究

传统的聚类方法能够自动实现类型的划分,但其只考虑空间对象的属性特征,忽略了空间对象之间位置的邻近性。在尺度空间理论基础上,利用多尺度空间聚类方法,同时兼顾空间对象间的空间相邻和属性相似这两个特征,使得在空间尺度由小变大的过程中,具有高度空间相互依赖关系的空间对象相互融合,得到不同空间尺度下的区域划分结果。并以福建省67个县市1990—2004年历年的人均GDP数据分析为例,进行了全省的经济区域划分应用试验。结果表明与实际的区域经济发展情况相符。该方法能够综合考虑空间位置邻近性和属性特征相似性的聚类要求,使得聚类结果更加合理,为制定区域经济持续发展战略提供依据。

A Study of Economic Regionalization Based on Multiscale Spatial Clustering

Automatic area division can be carried out by traditional cluster methods.But the attribute character was only considered in these methods,and the proximity between spatial objects was ignored.The method proposed using multi-scale spatial cluster was based on the theory of scale space.The proximity and attribute characters were combined in this method.So the great interdependent spatial objects were merged,and the division results with different spatial scale were obtained.The multiscale spatial clustering ...