摘要： 由于网上信息的迅速增长，文本分类已经成为网上信息管理中的热点研究课题．现今流行方法的重要基础是传统的统计学，前提是要有足够的样本，当样本数目有限时容易出现过学习的问题，导致分类效果不理想。本文首先从理论上分析了这种出现过学习的原因，然后介绍了出现过学习现象的条件和解决办法，最后介绍了一种研究小样本情况下机器学习规律、可以较好地克服过学习问题的分类方法--支持向量机方法。

关键词: 过学习, 支持向量机, 自动分类

Abstract: With the rapid growth of network information, text categorization has become an important research field for the management of network information. Most of the existing methods are based on traditional statistica which provides that conclusion only for the situation where sample size is tending to infinity. So they may not work well in practical case with limited samples and easily lead to the problem of overfitting. This paper theoretically analyze the reason of overfitting and introduce its condition as weIl as the method to solve it. At last we introduce SVMs, a method to avoid overfitting, which is based on statistical learning theory matching the limited number of examples.

中图分类号: 

• TP393.01