摘要
为了提高蚁群化学聚类算法的聚类效果,通过引入径向基核函数改进蚁群化学聚类算法的相似度度量式,提出了核蚁群化学聚类算法。将核蚁群化学聚类算法用于三个标准数据集合,计算结果表明与蚁群化学聚类算法相比,核蚁群化学聚类算法聚类效果提升显著。将核蚁群化学聚类算法、核自组织神经网络映射算法和基于多项式核的结构化有向树数据聚类算法同时用于Iris数据集合,结果显示三种核聚类算法聚类效果相当。
This paper introduced a new method—kernel AntClust(KANTCLUST) by using the radius kernel function to change the similarity function of ANTCLUST in order to improve the effect of ANTCLUST.It used KANTCLUST in three standard data set.The experiment results show that the clustering effect of KANTCLUST is better than that of ANTCLUST.KANTCLUST compared with kernerl self-organizing maps (KSOM) and polynomial kernel based structural clustering algorithm by building directed trees(SDTC).They were used in Iris standard data set.The result shows the clustering effect of KANTCLUST is as good as the classifying effect of KSOM and the clustering effect of SDTC.
出处
《计算机应用研究》
CSCD
北大核心
2010年第4期1326-1329,共4页
Application Research of Computers