文本分类中基于逆云模型的特征选择方法

Feature selection method based on backward cloud model in text classification

提出一种基于逆云模型的特征选择算法.该算法根据逆云模型理论建立训练集各属性在各类别上的模型,根据所建模型计算每个属性的类间差别,选取类间差别大的属性作为分类特征.该方法同时考虑了特征的频率因素.选择Reuter-21578和复旦中文语料数据集进行实验,并与信息增益、文本证据权重和互信息特征选择方法进行比较.结果表明,基于逆云模型的特征选择方法与信息增益方法在分类性能上大致相当,优于文本证据权重和互信息方法特征选择方法.

A feature selection method based on backward cloud model was proposed.The model of each feature in each class was expressed according to the theory of backward cloud model,and the distinction of each feature between different classes was calculated,The features with larger distinction between classes were selected.In addition,the frequency of the feature was considered.The feature selection method was applied to Reuter-21578 and Chinese text dataset provided by Fudan Database Center,and compared with information gain（IG） method,WET method and mutual information（MI） method.Experimental results show that the performance of the proposed feature selection method is comparable with that of IG method and higher than that of WET and MI methods.