基于两层主动学习策略的SVM分类方法  

SVM Classification Method Based on Two-Level Active Learning Strategy

作　　者：孟光胜[1] 赵志宇[1] 

机构地区：[1]河南科技大学林业职业学院,河南洛阳471002

出　　处：《河南师范大学学报（自然科学版）》2014年第2期158-162,共5页Journal of Henan Normal University(Natural Science Edition)

基　　金：河南省教育厅科技攻关项目(2010B520033)

摘　　要：针对当前主动学习策略直接用于支持向量机(SVM)分类器时存在泛化能力不强的问题,提出了两层主动学习策略(TLAC),该策略利用协调训练的思想,深层挖掘未标记样本数据的分布知识,从而选择最有利于分类器性能的样本来训练分类器.实验表明,该TLAC策略能够合理地指定TSVM算法中的正样本数,在典型指标测试中都表现出了一定的优越性.To deal with the poor generalization problem when active learning strategy directly using in SVM classifier, a two-level active learning strategy（TLAC） was proposed. By means of the idea of co-training, it deeply mines the distribution knowledge to select positive labeled samples which are most conducive to train a classifier. The experiment results show that TLAC strategy can determine the positive labeled sample numbers reasonable and demonstrate its superiority in typical indicator test.

关 键 词：主动学习 协同训练 贝叶斯网络 支持向量机 

分 类 号：TP315[自动化与计算机技术—计算机软件与理论]