一种基于改进信息增益特征选择的最大熵模型文本分类方法

针对传统信息增益(IG)特征选择算法忽略词频分布的缺陷,该文提出一种新的IG特征选择算法.该算法通过引入均衡比和类内词频位置参数,解决了传统IG算法忽略词频分布对分类的弱化问题,修正传统类内词频位置参数,提高特征选择算法的文本分类精度,并将该改进IG特征选择算法用于最大熵模型(ME)对文本进行分类.实验结果表明:该文所提方法在进行文本分类时F1值高于传统IG算法.该文方法的ME分类精度高于K最近邻KNN(K-Nearest Neighbor)算法,说明本文方法是可行的、有效的.

A Maximum Entropy Model Text Classification Method Based on Improved Information Gain Feature Selection

For the shortcomings of traditional information gain (IG) feature selection algorithm:ignoring word frequency distribution,a new IG feature selection algorithm is proposed in this paper.The algorithm introduces the equalization ratio and word frequency location parameters within class.The new algorithm solves the problem that the traditional IG algorithm ignores the word frequency distribution and modifies the position parameter of word frequency within class to improve the accuracy of text classification.At last,the improved IG feature selection algorithm is applied to maximum entropy model (ME) for text classification.Experimental results show that,Compared with the traditional IG algorithm,the F1 value of the proposed method in this paper is higher than the traditional IG algorithm in text classification.In addition,the ME classification accuracy of this method is higher than the KNN algorithm,which shows that this method is feasible and effective.