基于两层协同主动学习策略的分类算法  

Classification Algorithm Based on Two-Level Collaboration Active Learning Strategy

作　　者：张志锋[1] 范乃梅[1] 

机构地区：[1]郑州轻工业学院软件学院,河南郑州450002

出　　处：《青岛科技大学学报（自然科学版）》2013年第5期528-531,共4页Journal of Qingdao University of Science and Technology:Natural Science Edition

摘　　要：针对当前主动学习策略直接用于SVM(Support Vector Machine)分类器时存在的泛化能力不强的问题,提出的两层协同主动学习策略TLCALS(Two-Level Collaboration Active Learning Strategy)应用了协同训练的思想,能深层挖掘未标记样本数据的分布知识。实验表明,TLCALS策略能够合理指定TSVM(Transducive Support Vector Machine)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 is directly used in SVM classifier, a TLCALS （Two-Level Collaboration Active Learning Strategy） is proposed. It uses the idea of co-training, and can deeply mine the distribu- tion knowledge. The experiment results show that TLCALS strategy can determine the positive labeled sample numbers reasonable and demonstrate its superiority in typical in- dicator tests.

关 键 词：支持向量机 两层协同主动学习策略 贝叶斯网络 

分 类 号：TP315[自动化与计算机技术—计算机软件与理论]