文本多分类中的特征选择研究

Feature Selection for Multi-Class Text Categorization

特征选择是数据挖掘和机器学习等领域内重要的预处理步骤,近年来得到了广泛的关注。文本数据的高维性往往会影响分类等数据挖掘任务的效率,因此特征选择常被作为文本分类过程中的重要组成部分,以达到降维的目的。随着分类技术的快速发展,类别的日益细化,文本的多类分类问题为特征选择方法提出了更多的挑战。本文面向文本多类分类的应用背景,阐述了目前特征选择方法所面临的主要挑战,给出了多分类特征选择方法的主要种类。本文沿着相关研究的发展路线,由易至难,由浅入深,通过对目前多分类特征选择算法的应用情况进行总结,并进行综述评论,最后对全文进行了概括,提出了未来可能的研究方向。

As an important pre-processing step in data mining and machine learning, feature selection has been gradually developed. The high-dimensional characteristics of text data always declines the performance of categorization. Hence, feature selection can be employed as a dimension-reduction measure. With the fast evolution of classification methods and refinement of categories, muhi-class text categorization gives rise to more challenges for feature selection. In this paper, we present a survey of the main problems and the state of-art feature selection methods, following the development track. Finally, we conclude the whole paper and give some future directions of research.