支持向量机优化基于K-means的蚁群聚类算法  

SVM optimizing the K-means based antclust algorithm

作　　者：莫锦萍[1] 张志刚[2] 

机构地区：[1]广西财经学院现代教育技术部,广西南宁530003 [2]广西水利电力职业技术学院,广西南宁530024

出　　处：《微型机与应用》2012年第6期76-79,共4页Microcomputer & Its Applications

摘　　要：基于K-means算法思想改进蚁群聚类算法聚类规则,提出一种新的K-means蚁群聚类算法,并通过实验验证其聚类效果;引入具有全局最优性的支持向量机SVM,取各类中心附近适当数据训练支持向量机,然后利用已获模型对整个数据集进行重新分类,进一步优化聚类结果,使聚类结果达到全局最优。UCI数据集实验结果表明,新的算法可以明显提高聚类质量。Firstly this paper proposes an improved AntClust algorithm （KM-AntClust）, which optimizes the rules of AntClust model with K-means mind, then vertifying the clustering effection by experiments. Secondly introducing SVM to turther improve the clustering effection, In this step, the SVM is trained with dataset beyond clusters center at frist, then gaining the global optimal clusters when SVM is utilized to reclassify the original datasets. Experimental results for UCI datasets demonstrate that the improved method can obviously improve the classification quality.

关 键 词：K-平均算法 蚁群算法 聚类 支持向量机 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]