基于两层主动学习策略的SVM分类方法

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.

SVM Classification Method Based on Two-Level Active Learning Strategy

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.