基于改进特征选择方法的文本情感分类研究

提出了结合情感词典的改进信息增益特征选择方法。首先，针对现有的信息增益特征选择存在注重特征词的文档频率而忽视语料均衡等问题，提出了改进方法。其次，考虑情感词对文本分类的影响，提出了基于情感词典的特征选择（information gain combining sentiment classification，IGSC）算法进行文本分类。该算法通过对文本情感词进行匹配并结合情感词赋权重，实现了特征降维并解决了文本数据稀疏影响分类性能的问题；最后，针对旅游评论数据集对所提出的特征选择方法进行了实验验证及分析。实验结果表明，本文提出的改进文本情感分类特征选择方法在分类准确率、召回率和F值方面均得到了提升，并且具有较好的分类稳定性。

Research on text sentiment classification based on improved feature selection method

An improved information gain feature selection method based on sentiment dictionary was proposed.Firstly,aiming at the existing problems of information gain feature selection,such as paying attention to the frequency of feature word and ignoring the balance of corpus,an improved method was proposed.Secondly,considering the influence of sentiment words in text classification,a feature selection method IGSC (information gain combining sentiment classification) based on sentiment dictionary was proposed for text classification.By matching the text emotion words and combining the weight of emotion words,the feature dimension reduction was realized and the problem of text data sparseness affecting classification performance was solved.Finally,according to the proposed feature selection method of travel review data set for experimental verification and analysis,the experimental results show that the improved text sentiment classification feature selection method has been improved in terms of classification accuracy,recall and F value,and classification has better stability.