Multidimensional hierarchical classification
作者：
Highlights：
• Our proposal is a new hierarchical classifier inspired in multidimensional classification.
• We explored all the possible paths that an example can take in the hierarchy contrary to the traditional way.
• The method proposed avoids the inconsistency problem present in hierarchical top–down approaches.
• We propose an extension, based on information gain, to make predictions at any level of the hierarchy.
• Our proposal can be applied in tree or DAG hierarchies.
摘要
•Our proposal is a new hierarchical classifier inspired in multidimensional classification.•We explored all the possible paths that an example can take in the hierarchy contrary to the traditional way.•The method proposed avoids the inconsistency problem present in hierarchical top–down approaches.•We propose an extension, based on information gain, to make predictions at any level of the hierarchy.•Our proposal can be applied in tree or DAG hierarchies.
论文关键词：Hierarchical classification,Multidimensional classification,Information gain
论文评审过程：Available online 23 June 2014.
论文官网地址：https://doi.org/10.1016/j.eswa.2014.05.054
