摘要
Hilbert曲线能够线性填充数据空间,将数据空间分割成大小相等的网格,从而将位于网格中的点映射到线性空间中.本文利用Hilbert曲线的数据聚类性质,提出一种基于Hilbert曲线网格划分聚类算法,详细叙述算法的执行过程,并给出每一步的理论依据.算法首先以网格为单位合并出面积较小的聚集,然后将小聚集经过若干次合并形成较大聚集,最终使得聚集最优.实验结果表明该算法的执行时间少于经典聚类算法k-m eans和基于网格聚类算法CLIQUE.
Hilbert curve can fill the data space linearly,divide it into equal - size grids and map points lying in grids into the linear space. Using the quality of the clustering of Hilbert curve, the paper presents a gdd-partiton clustering algorithm based on Hilbert curve. First, the algorithm merges numerous small clusters based on the grids. Then it merges the clusters again and again. Finally, it gets large clusters. According to the test, the algorithm is better than the clustering algorithm k-means and CLIQUE.
出处
《小型微型计算机系统》
CSCD
北大核心
2010年第10期1979-1983,共5页
Journal of Chinese Computer Systems
基金
黑龙江省自然科学基金项目(F200601)资助
关键词
HILBERT曲线
网格划分
降维
聚类算法
Hilbert curve
grid partition
reduction of dimensionality
clustering algorithm