摘要
谱聚类算法利用特征向量构造简化的数据空间,在降低数据维数的同时,使得数据在子空间中的分布结构更加明显。现有谱聚类算法的聚类结果多为精确集,而真实数据集中重叠现象广泛存在。基于粗糙集理论提出了一种新的谱聚类算法,其主要思想是对谱聚类算法进行粗糙集扩展,使得聚类结果成为具有下近似和上近似定义的、类与类之间存在重叠区域的结构。实验表明,该算法与现有的谱聚类算法相比,稳定性和准确率都有一定的提高。
The spectral clustering algorithm constructs a simplified data space making the use of the eigenvectors that not only reduces the dimension of data but also gives clearer distribution of data in the subspace. The results of most existing spectral clustering algorithm are precise sets while widespread ＇ overlapping＇ exists in real data sets. This paper proposed a new spectral clustering algorithm which is based on the rough set theory. The main idea is to extend spectral clustering with rough set theory to obtain the results with lower-and-upper-approximation definition and between-cluster-overlapped structure. Experiment results indicate that the proposed algorithm outperforms the existing spectral clustering algorithms in both stability and accuracy.
出处
《计算机科学》
CSCD
北大核心
2009年第5期193-196,共4页
Computer Science
基金
国家自然科学基金(60475019
60775036)
教育部博士点专项基金(20060247039)资助
关键词
粗糙集
谱聚类
K均值聚类
Rough set, Spectral clustering, K-means clustering
