Data-driven global-ranking local feature selection methods for text categorization

作者：

Highlights：

• We propose two filtering methods for text categorization.

• The proposed methods find the number of features based on a low range parameter.

• The proposal ensures contribution of each document in the final subset of features.

• The proposed methods obtain better results than VR and ALOFT algorithms.

摘要

•We propose two filtering methods for text categorization.•The proposed methods find the number of features based on a low range parameter.•The proposal ensures contribution of each document in the final subset of features.•The proposed methods obtain better results than VR and ALOFT algorithms.

论文关键词：Text classification,High dimensionality,Feature selection,Filtering method,Variable Ranking,ALOFT

论文评审过程：Available online 16 October 2014.

论文官网地址：https://doi.org/10.1016/j.eswa.2014.10.011