一种快速山峰聚类算法*

山峰聚类既可以对数据集进行近似聚类，又可以为其他聚类方法提供聚类所需的初始聚类中心。减法聚类是山峰聚类的改进，它避免了山峰聚类中出现的计算量随样本维数增加呈指数增长的情况。但减法聚类对处理大样本集也力不从心。引入了P-tree数据结构，对高维大样本集进行分解，然后用减法聚类对子样本集进行聚类。此算法既避免了山峰聚类的维数灾难问题，也解决了减法聚类中样本数太大的问题。实验结果证明，该算法有效地减少了运算量，提高了聚类的速度。

Quick mountain clustering algorithm

A new clustering technique is described, which is an improvement on the mountain method (MM) of clustering originally.For higher dimensional data sets, the MM approach becomes computationally unattractive or even infeasible.Subtractive clustering method is an improvement on the mountain method. But for large data sets the SCM can still be computatio-nally intensive. This paper used P-tree data structure to decomposing the higher dimensional and large data sets, then clustered the small data sets using SCM.The method not only avoids the question of higher dimensional, but also solves the shor-tage of large data sets of SCM.