考虑标记间依赖关系的多标记分类算法

Multi-label Classifier Using the Dependency among Labels

  • 摘要: 提出了一种考虑标记间依赖关系的多标记分类算法.首先依据 RAkEL 算法将标记集合划分为若干子集,然后在子集内部应用概率分类器链算法训练分类器.这样不仅充分考虑了标记间的依赖关系,而且对标记进行分组,从而提高了分类的性能.在 5 个数据集上与其他经典算法进行了对比实验,结果表明本文所提算法可显著提高分类性能.

     

    Abstract: A multi-label classifier considering the dependency among labels was proposed. The label set was divided into several subsets based on RA k EL algorithm. Then, a classifier was built on each subset using probabilistic classifier chain. By experimental evaluations on five datasets, the results showed that the proposed algorithm achieved better performance than previous methods.

     

/

返回文章
返回