一种可信子空间标志方法*

在D-S证据理论的基础上,给出了可信子空间的定义及能够发现所有可信子空间的贪心算法CSL(creditable subspace labeling)。该方法迭代地发现原始特征空间的信任子空间集Cs。用户根据应用领域的需求, 对Cs中的每个可信子空间调用传统聚类算法发现聚类结果。实验结果表明,CSL具有正确发现原始特征空间的真实子空间的能力,为传统聚类算法处理高维数据空间聚类问题提供了一种新的途径。

Creditable subspace labeling method

This paper first gave the definition of creditable subspace based on D-S evidence, and then proposed a greedy algorithm CSL(creditable subspace labeling)which could search all the creditable subspaces.The creditable subspaces of the original data space could be found by iteratively executing the algorithm CSL.Once the creditable subspaces Cs were got, the true clustering results could be found using a traditional clustering algorithm on each creditable subspace. Experiments show that CSL can detect the actual creditable subspace with the original attribute. In this way,proposed a novel approach of using traditional clustering algorithms to deal with high dimensional data set clustering problems.