Mutual Information-based multi-label feature selection using interaction information
作者：
Highlights：
• We proposed a MI-based feature selection method without problem transformation.
• A score function measuring dependency between features and labels was derived.
• We derived theoretical bounds of score function.
• Based on theoretical bounds, a score function of variations from MI was chosen.
摘要
•We proposed a MI-based feature selection method without problem transformation.•A score function measuring dependency between features and labels was derived.•We derived theoretical bounds of score function.•Based on theoretical bounds, a score function of variations from MI was chosen.
论文关键词：Multi-label feature selection,Multivariate feature selection,Interaction information,Feature dependency
论文评审过程：Available online 18 October 2014.
论文官网地址：https://doi.org/10.1016/j.eswa.2014.09.063
