SVM文本分类中一种新的特征提取方法

A New Feature Selection Method in SVM Text Categorization

随着互联网的迅速发展,面向重要网络媒体海量发布信息实现智能分类,对于网络信息监管、舆论引导工作有着深远的意义。文中针对在文本分类中的特征选取问题,描述了一种基于法矢量权重的特征评价和选取方法。将此方法与SVM学习算法进行结合,在路透社标准文本测试集上进行了对比评估。实验结果显示,此特征选取方法相对于传统的特征选取方法可以产生更优的分类性能。此特征提取方法提供一种有效的途径,在基本保持分类器性能的前提下显著地减少特征空间的维数,进而提升系统的资源利用效率。

With the rapid development of Intemet, it has momentous significance for the task of the surveillance and management of network and leading the public to carry out the intelligence classification of the massive amount of information that released by the important network medium. This paper describes a feature selection method based on the weight of normal from SVM model. Using this feature scoring method with SVM learning algorithm on standard Reuters test set to compare other traditional feature selection method： Odds Ra- tio, Information Gain. Experimental results show that the normal weight based method yield better classification performance. This feature selection method provides an effective way to maintain the classification performance while reducing the dimension of feature space and significantly enhances the efficiency of computing resources.