关键词: 聚类分析, 树型索引, 网格划分, Hilbert R-tree索引

Abstract: In the traditional clustering methods based on grid and density，the interval and the density valve needs to be input with unsmoothed cluster，wrong judgement of clustering boundary.A clustering approach is proposed，which can confirm the interval and the number of clustering，reasonable cluster the object and create the Hilbert R-tree using the result of clustering.Experiment results show that the method is better than traditional Hilbert R-tree index in building time and other aspects.

Key words: clustering analysis, tree-like index, grid partition, Hilbert R-tree index