An Iterative Decision Tree Threshold Filter

Oscar Picchi NettoJosé Augusto Baranauskas

In this paper we propose and analyze a new filter for feature subset selection using an iterative decision tree threshold method. Using several biomedical or bioinformatics datasets, the filter has been evaluated on its data compression ability and AUC (Area Under Curve) performance within three scenarios. On average, the filter compressed almost 50% of the data. Additionally, AUC values using all versus selected filter features have not produced performance degradation in five different machine learning algorithms.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: