一种基于标签关联性的分层分类共有与固有特征选择方法

A Method of Common and Inherent Features Selection of Hierarchical Classification Based on Label Relevance

教学资源数据样本中,数据量大,且维度也逐渐增大,样本类别间存在层次结构,且不同类节点具有共有与固有属性的特点,导致其特征选择的难度较大,针对教学数据的特征选择方法无法充分利用类别的层次结构信息。为实现分层分类教学中层次结构数据的高效管理,降低特征选择的失败率,提出了一种基于标签关联性的分层分类共有与固有特征选择新方法。实验表明,特征选择失败率可始终控制在10%以内,该方法可行、高效。

In the teaching resource data samples,the amount of data is large,and the dimensions are gradually increasing.There are hierarchical structures among sample categories,and different kinds of nodes have the characteristics of common and inherent attributes,which lead to the difficulty of feature selection.The feature selection method for teaching data cannot make full use of the hierarchical structure information of categories.In order to realize the efficient management of hierarchical structure data in hierarchical classification teaching and reduce the failure rate of feature selection,a new method of common and inherent feature selection of hierarchical classification based on label relevance is proposed.The experiment shows that the failure rate of feature selection can always be controlled within 10%,and this method is feasible and efficient.