二分网格的多密度聚类算法

利用单元间的数据分布特征,提出了二分网格的多密度聚类算法BGMC.该算法根据两相邻单元的相邻区域中样本数量的积比两相邻单元的数据量积的相对数,判断两单元间的关系,寻找相似单元和边界单元,确定边界单元数据归属.实验结果表明,该算法可以很好的区分不同密度、形状和大小的类,聚类结果与数据输入顺序和起始单元选择顺序无关,算法执行效率高,具有良好的空间和维数的可扩展性.

Bisecting grid-based multi-density clustering algorithm

A bisecting grid-based multi-density clustering algorithm(BGMC)using data distribution characteristics within units is proposed.This algorithm can estimate relationship of two units according to the relative number,which is calculated by dividing the products of data quantities from the two neighboring parts by the products of data quantities from two adjacent units after bisected the two units.Based on the relationship,it can seek for the similar unit and boundary unit to define the data attribute in boundary unit.The experimental results show that the algorithm can differentiate the different cluster with arbitrary shape and multi-density data sets effectively.The clustering results have no significant relationship with data input and the order of the units.Also it has a good data scale and the extendibility of dimension.