摘要
针对聚类中不规格形状数据点分布的处理难题,提出了一种基于密度梯度的聚类算法(CDG)。算法通过分析数据样本及其周边的点密度变化情况,选择沿密度变化大的方向寻找不动点,从而获取原始聚类中心,再利用类间边界点的分布情况对小类进行合并。实验结果表明,新算法较基于密度的带噪声数据应用的空间聚类方法(DBSCAN)具有更好的聚类性能。
In order to solve ditticult problems in clustering with irregularly distributed data set, a new clustering algorithm based on density gradient was provided. By analyzing the changing density of data sample and its neighbors, the algorithm searched points with the maximum density and took them as original centers of clusters. Then it combined some smaller clusters into larger ones according to the distribution of border points between clusters. Experimental results show that the new algorithm has better performance than Density Based Spatial Clustering of Applications with Noise（DBSCAN）.
出处
《计算机应用》
CSCD
北大核心
2006年第10期2389-2392,2404,共5页
journal of Computer Applications
基金
福建省自然科学基金资助项目(A0510024)
福建省青年基金(2005J051)
广东省关键领域重点突破项目(2005A10207003)
关键词
聚类
模式分类
数据挖掘
clustering
pattern classification
data mining
