Optimizing F-Measures by Cost-Sensitive Classification - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

Optimizing F-Measures by Cost-Sensitive Classification

Résumé

We present a theoretical analysis of F-measures for binary, multiclass and mul-tilabel classification. These performance measures are non-linear, but in many scenarios they are pseudo-linear functions of the per-class false negative/false positive rate. Based on this observation, we present a general reduction of F-measure maximization to cost-sensitive classification with unknown costs. We then propose an algorithm with provable guarantees to obtain an approximately optimal classifier for the F-measure by solving a series of cost-sensitive classification problems. The strength of our analysis is to be valid on any dataset and any class of classifiers, extending the existing theoretical results on F-measures, which are asymptotic in nature. We present numerical experiments to illustrate the relative importance of cost asymmetry and thresholding when learning linear classifiers on various F-measure optimization tasks.
Fichier principal
Vignette du fichier
5508-optimizing-f-measures-by-cost-sensitive-classification.pdf (1.18 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01196627 , version 1 (10-09-2015)

Identifiants

  • HAL Id : hal-01196627 , version 1

Citer

Shameem Puthiya A. Parambath, Nicolas Usunier, Yves Grandvalet. Optimizing F-Measures by Cost-Sensitive Classification. Advances in Neural Information Processing Systems 27, Dec 2014, Montréal, Canada. ⟨hal-01196627⟩
518 Consultations
683 Téléchargements

Partager

Gmail Facebook X LinkedIn More