摘要： 在传统网格密度方法的基础上，提出了一种新型的面向GIS 系统的双层聚类算法CPDG。它结合了密度方法和划分方法的思想，能够准确、高效地同时发现两种性质的空间集簇。初步实验显示，该算法在同时发现空间对象的任意形状密集连通区域以及连通区域内部以密集中心为代表的若干近似圆形的子区域上有很好的效果。

关键词: 聚类；空间聚类；基于网格；GIS

Abstract: This paper presents a new approach, which is developed from traditional grid density-based methods and combines the ideas of density-based methods and partitioning methods. New methods can exactly and efficiently divide the objects in geography space into some density-based regions and divide each of these regions into some sub-regions with representative center at the same time

Key words: Clustering; Spatial clustering; Grid-based; GIS