基于改进Canopy的FCM聚类算法
FCM Clustering Algorithm Based on Improved Canopy
孙偲远 1马鑫海1
作者信息
- 1. 江苏科技大学计算机学院 镇江 212100
- 折叠
摘要
传统FCM聚类算法对初始聚类中心较为依赖,针对其随机选取初始聚类中心的特点很容易造成目标函数陷入局部最优解,且对于大规模数据集收敛速度较慢等问题,论文提出了基于密度的Canopy算法对原始的FCM算法加以改进实现对初始聚类中心自动选取,将选取好的聚类中心作为FCM算法输入,以此来加快算法收敛并降低随机选取对分类结果造成的精度不准确的问题.
Abstract
The traditional FCM clustering algorithm relies on the initial clustering center.According to the characteristics of randomly selecting the initial clustering center,it is easy to cause the objective function to fall into the local optimal solution,and the convergence speed of large-scale data sets is slow.This paper proposes a density based canopy algorithm to improve the original FCM algorithm,so as to automatically select the initial clustering center,and take the selected clustering center as the input of FCM algorithm,in order to speed up the convergence of the algorithm and reduce the inaccurate accuracy of the classification re-sults caused by random selection.
关键词
聚类算法/聚类中心/FCM算法/密度Canopy算法Key words
clustering algorithm/cluster center/FCM algorithm/density Canopy algorithm引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据