基于多值分解和多类标学习的分类框架设计

Framework of Classification Based on Multi-Value Decomposition and Multi-Label Learning

多值多类标的数据分类是研究一个样本不但同时属于多个类别,而且在某些属性下也可能存在多个取值的问题。提出了一种结合多值分解和多类标学习的多值多类标分类框架(MDML),采用4种不同的多值分解策略,将问题转化为多类标问题,然后利用3种经典的多类标算法进行学习。实验结果表明,MDML与已有的多值多类标决策树算法相比,有效地提高了分类的性能,而且不同的组合方法适用于不同特点的数据集。

Classification of multi-valued and multi-labeled data is about a sample which is not only associated with a set of labels,but also with several values that include some attributes.This paper proposes a multi-valued and multi-labeled learning framework that combines multi-value decomposition with multi-label learning（MDML）,using four strategies to deal with multi-valued attributes and three classical,multi-label algorithms to learn.Experimental results demonstrate that MDML significantly outperforms the decision tree based method.Meanwhile,combined methods can be applied to various types of datasets.