摘要
通过研究核聚类算法,以及粗糙集,提出了一个新的用于聚类分析的粗糙核聚类方法。通过mercer核映射把输入空间中的样本映射到Hilbert空间,使样本空间中没有显现的特征在特征空间中突现出来,在这种样本差异加大的基础上,结合粗糙集的思想,把样本分别划到相应聚类中心的上、下近似中,上、下近似中的样本按照一定的比例来共同决定新的聚类中心。这样不但聚类精度大大提高,而且算法收敛速度也较快。仿真实验的结果表明该算法的可行性和有效性。
By means of analyzing kernel clustering algorithm and rough set theory, a novel clustering algorithm, Rough kernel k-means clustering algorithm, was proposed for clustering analysis. Through using Mercer kernel functions, samples in the original space were mapped into a high-dimensional feature space, which the difference among these samples in sample space was strengthened through kernel mapping, combining rough set with k-means to cluster in feature space. These samples were assigned into up-approximation or low-approximation of corresponding Clustering centers, and then these data that were in up-approximation and low-approximation were combined and to update cluster center. Through this method, clustering precision was improved, clustering convergence speed was fast compared with classical clustering algorithms. The results of simulation experiments show the feasibility and effectiveness of the kernel clustering algorithm.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2008年第4期921-925,共5页
Journal of System Simulation
基金
国家自然科学基金(60472072)
陕西理工学院科研项目(SLG0631)
