A classification rule is performed to assign a class to new sample indi- viduals. Many times the number of classes is large, consequently the classification rule could have some problem in reaching a satisfying level of accuracy. We deal with an approach called Sequential Automatic Search of a Subset of Classifiers (SASSC), able to enhance the classification rule performance and the interpretability of its output. It consists in splitting a classification problem among C classes into K < C less complex two-classes sub-problems and evaluate its performance on two different datasets. The main contributions of SASSC concern the new criteria for the aggregation of classes and super-classes and the alternative criteria for the estimation of the response class for unseen (test-set) observation.

Combined methods in multi-label classification algorithms

Frigau Luca
;
Conversano Claudio;Mola Francesco
2017-01-01

Abstract

A classification rule is performed to assign a class to new sample indi- viduals. Many times the number of classes is large, consequently the classification rule could have some problem in reaching a satisfying level of accuracy. We deal with an approach called Sequential Automatic Search of a Subset of Classifiers (SASSC), able to enhance the classification rule performance and the interpretability of its output. It consists in splitting a classification problem among C classes into K < C less complex two-classes sub-problems and evaluate its performance on two different datasets. The main contributions of SASSC concern the new criteria for the aggregation of classes and super-classes and the alternative criteria for the estimation of the response class for unseen (test-set) observation.
2017
9788899459710
SASSC; Super-class; Classification; Multi-class
File in questo prodotto:
File Dimensione Formato  
CLADAG_2017_paper_144.pdf

Solo gestori archivio

Tipologia: versione editoriale
Dimensione 315.32 kB
Formato Adobe PDF
315.32 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11584/232061
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact