%0 Journal Article
%T 基于链分解的多标签分类属性约简<br>Attribute Reduction for Multi-Label Classification Based on Chain Decomposition
%A 张莹
%J Hans Journal of Data Mining
%P 240-246
%@ 2163-1468
%D 2020
%I Hans Publishing
%R 10.12677/HJDM.2020.104025
%X 本文提出了基于链分解的多标签属性约简方法。通过考虑标签之间的相关性，将标签进行排序，根据排序方法，多标签问题被分解成单标签链的形式，对于链中每一个子问题通过粗糙集方法重新定义下近似、正域、依赖度，并进行属性约简。实验结果表明，该方法能在不降低分类精度的情况下去除大部分冗余属性。<br />
In this paper, a new multi-label attribute reduction algorithm based on the chain decomposition is proposed. Considering the correlation between the labels, the labels are sorted. According to the sorting method, the multi-label problem is decomposed into a single-label problem chain. For each sub-problem, the lower approximation, the positive region and the dependency are redefined by the rough set method, and the attributes are reduced. Experimental results show that the algo-rithm can remove most of the redundant attributes without reducing the classification accuracy.
%K 多标签分类，属性约简，粗糙集，链分解<br>Attribute Reduction
%K Rough Set
%K Multi-Label Classification
%K Chain Decomposition
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=37859