基于自适应的粗糙C-均值聚类算法  

Rough C-means Clustering Algorithm Based on Self-adaption

作　　者：严静静[1] 张腾飞[1] 

机构地区：[1]南京邮电大学自动化学院,江苏南京210023

出　　处：《计算机技术与发展》2016年第3期67-70,共4页Computer Technology and Development

基　　金：江苏省普通高校研究生科研创新计划项目(46888LX14819)

摘　　要：粗糙C-均值的提出,首次将粗糙集与聚类算法结合起来。随后,众多学者对其进行了广泛研究。然而,绝大多数算法在研究簇的下近似、边界对象时,使用统一的权重,忽略了这些对象本身的差异性以及对所在簇的贡献。针对此问题,文中提出一种改进的聚类方法。通过样本对象偏移其所在簇心的程度,设定不同的簇偏移量,距离簇心越近的样本对象其簇偏移量越大,反之越小。通过此举以客观描述这些样本对象对其所在簇的贡献,使得最终聚类结果更加精确、簇内更加紧密、簇间更加稀疏。实例计算结果以及通过MATLAB对数据库中IRIS的数据集进行仿真验证,表明提出的改进算法具有一定的可行性。Rough C -means is proposed to combine the rough set with clustering algorithm first. In the following, many scholars have been doing extensive research. However,for the objects in the low approximation or boundary ,the most of algorithms use unified weights,ignoting the difference of the objects themselves and the contribution to the classes. Aiming at this problem,an improved clustering method is put forward. Based on degree of objects deviated centroid of clusters,it sets different offsets to highlight these objects on contribution to the classes in this paper,making the result of clustering more precise,intra-classes more close,and inter-classes more sparse. The experimental results and simulation verification on IRIS by MATLAB shows the method is feasible.

关 键 词：聚类 粗糙集 粗糙C-均值 簇偏移量 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]