基于学习自动机的改进FCM聚类算法及应用  

Improved FCM clustering algorithm and application based on learning automata

作　　者：张晋义 李凤莲 张雪英 黄丽霞 陈桂军 ZHANG Jinyi;LI Fenglian;ZHANG Xueying;HUANG Lixia;CHEN Guijun(School of Information and Computer,Taiyuan University of Technology,Jinzhong 030600,China)

机构地区：[1]太原理工大学信息与计算机学院,山西晋中030600

出　　处：《电子设计工程》2023年第9期13-18,共6页Electronic Design Engineering

基　　金：国家自然科学基金资助项目(62171307);山西省科技重大专项(20181102008)。

摘　　要：模糊C均值(Fuzzy C-Means,FCM)聚类算法存在过度依赖初始聚类中心,且未充分考虑隶属度矩阵变化趋势对聚类性能影响的缺陷,针对FCM存在的问题,提出了一种基于学习自动机的改进FCM聚类算法。算法改进了目标函数计算方式及隶属度矩阵,根据目标函数值以及平均类内距离的变化对智能体选择的行为进行奖励或者惩罚。通过UCI公共数据集以及工业生产中碳碳复合材料沉积炉生产数据进行实验,实验结果表明,相比K-means、FCM、IEWLFCM、LAC等几种已有聚类算法,在文中采用的实验数据集中,该文提出的基于学习自动机的改进FCM聚类算法在大多数数据集上准确率、FMI系数、JC系数均有所提升。Fuzzy C-means(FCM)clustering algorithm relies too much on the initial clustering center and does not fully consider the influence of membership matrix change trend on clustering performance.Aiming at the problems existing in FCM,an improved FCM clustering algorithm based on learning automata was proposed.The algorithm improves the calculation method of objective function and membership matrix.Reward or punish the agent′s chosen behavior according to the value of the objective function and the change of the average in⁃class distance.Experiments were carried out through UCI public data set and production data of carbon⁃carbon composite deposition furnace in industrial production.The experimental results show that,compared with K-means,FCM,IEWLFCM,LAC and other existing clustering algorithms,for the experimental data set adopted in this paper,the accuracy,FMI coefficient and JC coefficient of the improved FCM clustering algorithm based on learning automata proposed in this paper have been improved in most data sets.

关 键 词：FCM 学习自动机 奖惩机制 隶属度矩阵 聚类效果 

分 类 号：TN98[电子电信—信息与通信工程]