摘要
针对信息增益特征选择方法没有很好考虑多标记的分布,在多标记文本分类中表现不佳的问题,用标记矩阵的协方差改善特征选择时标记之间的关联产生的影响,提高分类效果。最后通过实验证明,改进的信息增益特征选择方法具有可行性和有效性。
To solve the problem of the poor effect of information gain feature selection on the multi-labeled text categorization,which does not consider multi labeled distribution.Class corelation is taken into account and use multi-label covariance to improve IG feature selection in classification.The experimental results verify the efficiency and probability of the improved information gain feature selection in Multi-labeled text categorization.
出处
《廊坊师范学院学报（自然科学版）》
2012年第5期46-48,共3页
Journal of Langfang Normal University(Natural Science Edition)
基金
莆田市科技项目[2011G04(2)]支持
关键词
文本分类
多标记分类
信息增益
特征选择
协方差
text categorization
Multi-labeled classification
information gain
feature selection
Covariance
