有向无环图-双支持向量机的多类分类方法

A DIRECTED ACYCLIC GRAPH AND TWIN SUPPORT VECTOR MACHINE ALGORITHM FOR MULTI-CLASS CLASSIFICATION

针对多分类支持向量机算法中的低效问题和样本不平衡问题,提出一种有向无环图-双支持向量机DAG-TWSVM(directed acyclic graph and twin support vector machine)的多分类方法。该算法综合了双支持向量机和有向无环图支持向量机的优势,使其不仅能够得到较好的分类精度,同时还能够大大缩减训练时间。在处理较大规模数据集多分类问题时,其时间优势更为突出。采用UCI(University of California Irvine)机器学习数据库和Statlog数据库对该算法进行验证,实验结果表明,有向无环图-双支持向量机多分类方法在训练时间上较其他多分类支持向量机大大缩短,且在样本不平衡时的分类性能要优于其他多分类支持向量机,同时解决了经典支持向量机一对一多分类算法可能存在的不可分区域问题。

In order to deal with the inefficient and sample unbalance problems in multi-classification support vector machine algorithm, we proposed a multi-classification method with directed acyclic graph and twin support vector machine （DAG-TWSVM）. This algorithm combines the advantages of twin support vector machine （TWSVM） and directed acyelic graph support vector machine （ DAG-SVM）, makes it get better classification accuracy as well as greatly reduces the training time simultaneously. When dealing with multi-classification of large scale dataset, its advantage in time is more prominent. To verify the proposed algorithm with UCI （university of California Irvine） machine learning database and Statlog database, experimental results showed that the DAG-TWSVM multi-classification method uses much less time in training than other multi-classification SVMs, and meanwhile it solves the problem of inseparable region possibly existed in one-to-one multi- classification of classic SVM.