支持向量聚类算法的研究与改进

Study and Improvement on the Support Vector Clustering Algorithm

支持向量聚类是基于支持向量机和核方法的一种新颖的聚类方法.与其它传统聚类方法相比较,该方法具有能得到全局最优解,并能处理任意形状的聚类,无需指定聚类数目,参数少,容易处理高维数据等优点.在原算法的基础上,在聚类标识阶段提出了改进算法,用支持向量代替原来的全部样本数据来进行标识,进一步减少运算时间,提高运算速度.

Support vector clustering(SVC) is a novel clustering method inspired by support vector machines(SVM) and kernel methods.By comparing SVC with the other traditional clustering methods,we find SVC has many advantages,such as a global optimum,treatment of data sets of arbitrary shape,no need for specifying the number of clusters,fewer parameters,and easy treatment of high dimensional data.This paper improves the traditional support vector clustering algorithm during the cluster assignment phase,using support vector instead of all samples of data sets to shorten the operational time and raise the operation speed.