基于支持向量数据描述的无标签数据多类分类  

Multi-class classification algorithm for unlabeled data using SVDD

作　　者：朱帮助[1] 林健[2] 

机构地区：[1]五邑大学系统科学与技术研究所,广东江门529020 [2]北京航空航天大学经济管理学院,北京100083

出　　处：《智能系统学报》2009年第2期131-136,共6页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金资助项目(70471074)

摘　　要：为解决支持向量机(SVM)在处理无标签数据多类分类上的难题,提出了一种基于支持向量数据描述(SVDD)的无标签数据多类分类算法.该方法只需要建立一个分类模型就可以实现多类聚类分类.首先采用主成分分析作数据预处理,提取输入数据的统计特征值,得到主成分特征指标输入到SVDD分类器进行多类聚类分类.以珠三角地区物流中心城市分类评价为研究对象,实证结果表明,采用主成分分析降低了数据维度,有效浓缩了评估信息,SVDD分类器很好地区分了各中心城市,实现了多类分类的目的.Support vector machines （SVM） may encounter problems in dealing with multi-class classification of unlabeled data. So we suggested a new multi-class classification algorithm based on support vector data description （SVDD） in this paper. Compared with other multi-class classification algorithms, the proposed algorithm only needed one classifier to complete the multi-class clustering classification. With this method, principal component analysis （PCA） was used to preprocess original data to extract statistically characteristic values; inputting these values into an SVDD classifier completed multi-class clustering classification. Taking nine cities in the Pearl River delta area as an example, an evaluation was made of the developmental levels of the logistics of these cities. The test results showed that data dimensions were reduced by using principal component analysis, and the evaluated information was effectively concentrated by adopting feature extraction with PCA. Moreover, the SVDD classifier could distinguish the central cities very well, so it can be used as an effective approach for multi-class classification of unlabeled data.

关 键 词：多类分类 无标签数据 支持向量数据描述 主成分分析 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]