New feature selection method for multi-class data: Iteratively weighted AUC (IWA)
Petr Honzik, Pavel Kučera, Ondřej Hynčica, Daniel Haupt · 2011
This paper deals with the new filter feature selection method Iteratively Weighted Area under Receiver Operating Characteristic (IWA). It is aimed for the multi-class problems with quantitative inputs. The experiments prove its superior quality in comparison to the equivalent methods.