基于密度加权的粗糙K-均值聚类改进算法  

Improved Rough K-means Clustering Algorithm with Weight Based on Density

作　　者：郑超[1] 苗夺谦[1] 王睿智[1] 

机构地区：[1]同济大学计算科学与技术系,上海201804

出　　处：《计算机科学》2009年第3期220-222,共3页Computer Science

基　　金：国家自然科学基金项目(60775036;60475019);高等学校博士学科点专项科研基金(20060247039)资助

摘　　要：针对粗糙K-均值聚类算法中类均值计算式的特点,提出了一种改进的粗糙K-均值算法。改进后的算法基于数据对象所在区域的密度,在类的均值计算过程中对每个对象赋以不同的权重。不同测试数据集的实验结果表明,改进后的粗糙K-均值算法提高了聚类的准确性,降低了迭代次数,并且可以有效地减小孤立点对聚类的影响。According to the feature of the calculation of means in Rough K-means algorithm, an improved Rough K- means algorithm was proposed. The new algorithm introduces weights to the calculation of means, which is based on the density of each point. The experiments show that the new algorithm improves the clustering accuracy and reduces the iteration times as well as the outliers＇ influence.

关 键 词：聚类算法 粗糙K-均值 密度 孤立点 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]