Optimizing F-measures by Cost-sensitive Classification

Puthiya Parambath, S. , Usunier, N. and Grandvalet, Y. (2014) Optimizing F-measures by Cost-sensitive Classification. In: Advances in Neural Information Processing Systems 27 (NIPS 2014), Montreal, Canada, 8-13 Dec 2014, pp. 2123-2131.

Full text not currently available from Enlighten.

Publisher's URL: https://papers.nips.cc/paper/2014/hash/678a1491514b7f1006d605e9161946b1-Abstract.html


We present a theoretical analysis of F-measures for binary, multiclass and multilabel 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.

Item Type:Conference Proceedings
Glasgow Author(s) Enlighten ID:Puthiya Parambath, Dr Sham
Authors: Puthiya Parambath, S., Usunier, N., and Grandvalet, Y.
College/School:College of Science and Engineering > School of Computing Science
Journal Name:Advances in Neural Information Processing Systems

University Staff: Request a correction | Enlighten Editors: Update this record