摘要
提出一种基于共享最近邻聚类和模糊集理论的分类器.首先,在提出与核点密切相关的核半径概念的基础上,应用共享最近邻聚类得到正常类空间的部分核点和核半径,建立求解正常类空间补充核点的多目标优化模型,从而获得刻画正常类空间的全部核点和核半径.然后,将模糊集理论引入正常类的类属划分中,利用核点和核半径定义正常类的隶属度函数,建立基于隶属度函数的分类函数或分类器.实验表明,该分类器能处理包含噪音、孤立点和不规则子类的高维数据集的分类问题.
A classifier based on shared nearest neighbor clustering and fuzzy set theory（SNNFT） is proposed. The concept of core radii closely related with core point is defined and a portion of core points and core radii are obtained by applying shared neighbor clustering. The multi-objective optimization model of complementary core points is established. Consequently all core points and core radii are obtained to depict normal class space. By introducing fuzzy set theory to partition normal class space, classification function or classifier based on membership function of normal class defined using core points and core radii are constructed. Experiments show that SNNFT can cope with classification problem with high dimension dataset which contains noise, outliers and irregular sub clusters.
出处
《控制与决策》
EI
CSCD
北大核心
2006年第10期1103-1108,共6页
Control and Decision
基金
国家自然科学基金重大研究计划项目(90104005)
国家重大基础研究前期研究专项(2003CCA00200)
关键词
分类器
共享最近邻聚类
模糊集
遗传算法
优化模型
Classifiers Shared nearest neighbor clustering
Fuzzy sets Genetic algorithms Optimization model
