摘要
通过对数据空间进行网格划分并寻找稀疏区域来发现类的边界，提出了一种基于密度与网格的新的子空间聚类算法．该算法使用投影寻踪的搜索策略来发现存在于子空间内的类，同时运用基于竞争的修剪方式来有效地控制算法的计算复杂性．实验结果表明，所提算法在精度、时间复杂性等方面具有优良性能.
A new kind of subspace clustering algorithm based on density and grids was proposed. Boundaries between classes are located by partitioning grids in the data space and finding sparse regions. The new algorithm uses the searching strategy of projected pursuit to find the classes in the subspace. A competitive pruning procedure is utilized to reduce the computational complexity. The experimental results on the benchmark datasets show that the proposed algorithm has advantages in accuracy and computational complexity.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2007年第4期577-577,共1页
Journal of Shanghai Jiaotong University
