Abstract: One-Class Support Vector Machine（OCSVM） is an important and widely used classifier. It can be used in the context in which the negative samples are hardly collected or labeled, such as intrusion detection, fault detection and diagnosis, and the classification of remote sensing data. Therefore, OCSVM has been attracting more and more attention on its theory research and applications in recent years. This paper reviews the two important algorithms for OCSVM, followed by the improvements on those algorithms, including using the unlabeled samples, selecting certain samples, and modifying the optimization problems. It makes conclusion for OCSVM.

Key words: Support Vector Machine, One-Class Support Vector Machine, classifier

摘要： 单类支持向量机是一种用途广泛的分类器，它能够应用于负类样本难以收集的领域中，如入侵检测、故障检测与诊断和遥感数据分类等领域。因此无论在理论研究还是实际应用方面，单类支持向量机受到越来越多的关注。回顾单类支持向量机的两种主要方法，阐述各种关于单类支持向量机的改进，包括使用未标号数据、选择样本点以及修改优化目标。对单类支持向量机做了总结。

关键词: 支持向量机, 单类支持向量机, 分类器