一种基于量子机制的分类属性数据模糊聚类算法

分类属性数据的样本间的分布不平衡、样本的分布与空间距离无关的特点与量子力学中粒子的分布状态由能量决定、粒子分布具有不平衡性的特点相似.基于此,参照量子聚类QC算法确定聚类中心的聚类策略,重写距离量子势能公式,定义相似性度量测度和相异性度量测度的新概念,提出了针对分类属性数据的量子聚类CQC算法,并对算法的聚类有效性进行了研究,通过同其它几个已有的算法的仿真实验比较,证明该算法是有效的、有一定的可扩展性,算法的一些性能优于已有的其它几个算法.

Fuzzy Clustering Algorithm for Categorical Data Using Quantum Mechanics

Categorical data are often unbalancedly distributed. Their distributions are often unrelated with their distance measure. These characteristics are very similar to the particle world in quantum mechanism,therefore,based on quantum clustering,a novel categorical quantum clustering CQC algorithm was proposed by rewriting the distance_based quantum potential equation and defining the new dissimilarity measure. Its validity was discussed. Several experimental results about categorical datasets demonstrate that the proposed algorithm outperforms the hard K-modes,fuzzy K-modes and conceptual K-means algorithms.