Abstract: CURE is a typical clustering algorithm that is designed for the mining of mass data.A new algorithm K－CURE is introduced in this article to improve the CURE based on partition.A method is also described to explain how to choose the occasion of eliminating outlier during clustering.Experiments indicate that the improved algorithm does improve the CURE in both efficiency and effectiveness．

摘要： CURE算法是针对大规模数据聚类算法的典型代表。提出了一种新的算法K－CURE,该方法基于划分思想对CURE算法作了改进,同时给出了在聚类中剔除孤立点的时机选择方法。测试表明,改进后的算法效率明显高于原算法,且聚类效果良好。