Abstract: Aiming?at?the?problem?that?the?SVM?multiclass?classification?algorithm?based?on?DAG?doesn＇t?use?an?effective?constructing?algorithm?of?directed?acyclic?graph,?an?improved?SVM?multiclass?classification?algorithm?based?on?DAG?is?put?forward.?The?class?distance?of?clustering?is?taken?as?the?basis?of?hierarchical?classification.?The?experiment?results?show?that?the?new?method?has?higher?classification?accuracy?than?the?original?algorithm?does.

Key words: supporting vector machines（SVG） , clustering , directed acyelic graph（DAG） , multiclass classification

摘要： 针对有向无环图支持向量机多类分类方法未采用有效的有向无环图生成算法,提出了一种改进的有向无环 图生成算法。该方法采用了聚类分析中类距离的思想作为层次分类依据。实验结果表明,该方法与原方法相比具 有较高的分类精度。

关键词: 支持向量机, 聚类, DAG, 多类分类