基于二次均值阴影集的FCM聚类算法

FCM Clustering Algorithm Utilizing the Quadratic Mean Shadow Set

闫珊珊 ¹李文焱 ¹李丽红²

扫码查看

作者信息

1. 华北理工大学理学院,河北唐山 063210
2. 华北理工大学理学院,河北唐山 063210;河北省数据科学与应用重点实验室,河北唐山 063210;唐山市数据科学重点实验室,河北唐山 063210
折叠

摘要

为减少五区域阴影集中使用常数值来定义阴影集的五值隶属函数而造成的划归误差,利用平均值隶属度构建了一种基于二次均值阴影集的FCM聚类算法(FCM-5QSS).首先,通过FCM算法得到对象与簇之间的隶属度矩阵;然后,使用平均值隶属度来定义阴影集的五值隶属函数,同时通过模糊度性质及不确定性平衡原理得到两对阈值,进而将隶属度对应的对象划分到 5 个区域中;最后,通过划分核心区域和次核心区域中隶属度µ≥η false的对象簇,得到聚类结果.在 8 个UCI公开数据集上的实验结果表明了所提出方法的有效性.

Abstract

In order to reduce the classification error caused by using constant values to define the five value membership function of shadow sets in five regions,a FCM clustering algorithm based on quadratic mean shadow sets(FCM-5QSS)was constructed using the average membership degree.Firstly,the membership matrix between objects and clusters is obtained through the FCM algorithm;Then,the average membership degree is used to define the five value membership function of the shadow set,and two pairs of thresholds are obtained through the properties of fuzziness and the principle of uncertainty balance.The objects corresponding to the membership degree are then divided into five regions;Finally,the clustering results are obtained by dividing the object clusters with membership degrees in the core and sub core regions.The experimental results on 8 UCI public datasets demonstrate the effectiveness of the proposed method.

关键词

三支决策/五区域阴影集/平均值隶属度/三支聚类

Key words

three branch decision-making/five region shadow set/average membership degree/three branch clustering

引用本文复制引用

出版年

2025

华北理工大学学报(自然科学版)

河北联合大学

华北理工大学学报(自然科学版)

影响因子：0.3

ISSN：2095-2716

段落导航