基于三支决策的密度敏感谱聚类

将三支决策与密度敏感谱聚类结合,提出了一种基于三支决策的密度敏感谱聚类算法。该算法通过在密度敏感谱聚类的聚类过程引入容差参数得到每个类的上界,然后通过扰动分析算法从上界中分离出核心域,上界和核心域的差值被认定为该类的边界域。聚类结果用核心域和边界域来表示每个类簇,可以更全面地展示数据的结构信息。与传统的硬聚类算法在UCI数据集的实验结果相比较,本文使用核心域计算聚类的评价指标DBI、AS和ACC都有所提升,较好地解决了不确定性对象的聚类问题。

Density-sensitive spectral clustering based on three-way decision

This paper integrates density sensitive spectral clustering with three-way decision and presents a density sensitive spectral clustering algorithm based on three-way decision. In the proposed algorithm, the upper bound of each cluster is obtained by introducing tolerance parameters in the process of density-sensitive spectral clustering, and the core is separated from the upper bound by the perturbation analysis algorithm. The difference between the upper bound and the core region is regarded as the fringe region of the specific cluster. The clustering result uses the core region and the fringe region to represent each cluster, which can more comprehensively display the data structure. Compared with the experimental results of the traditional hard clustering algorithm on the UCI data set, the proposed algorithm is effective in improving the value of AS and ACC and reducing the value of DBI by using the core region to calculate the evaluation indicators of clustering, which indicates that the proposed algorithm can be used to solve the problem of clustering uncertain objects.