基于GPU的文本特征选择与特征加权

通过对GPU通用计算与文本分类的研究,提出了一种基于GPU的文本特征选择与加权方法。首先介绍了文本分类中常用的特征选择方法和特征加权方法,并在GPU上实现了其中的DF(文档频率)方法和TFIDF方法。实验结果显示,利用提出的并行特征选择与加权方法能能有效的提高特征选择与加权过程的速度。

Text Feature Selection and Feature Weighting Based on GPU

Through the research on the GPU general computing and text categorization,this paper proposes a GPU-based text feature selection and weighting methods.This paper describes feature selection and feature weighting methods that the text classification method commonly used in and implemented DF （document frequency） method and TFIDF method on the GPU. Experiment result proves that the use of a parallel feature selection and weighting method in this paper,can effectively improve the feature selection and weighting process speed.