Online F-Measure Optimization
R´obert Busa-Fekete1, Bal´azs Sz¨or´enyi2, Krzysztof Dembczy´nski3, and Eyke H¨ullermeier1
1 Department of Computer Science, University of Paderborn, Warburger Str. 100, 33098 Paderborn, Germany
{busarobi,eyke}@upb.de
2 MTA-SZTE Research Group on Artificial Intelligence, Tisza Lajos krt. 103., H-6720 Szeged, Hungary
szorenyi@inf.u-szeged.hu
3 Institute of Computing Science, Pozna´n University of Technology, Piotrowo 2, 60-965 Pozna´n, PolandKrzysztof.Dembczynski@cs.put.poznan.pl
Abstract. 4 The F-measure is an important and commonly used per- formance metric for binary prediction tasks. By combining precision and recall into a single score, it avoids disadvantages of simple metrics like the error rate, especially in cases of imbalanced class distributions. The problem of optimizing the F-measure, that is, of developing learning algo- rithms that perform optimally in the sense of this measure, has recently been tackled by several authors. In this paper, we study the problem of F-measure maximization in the setting of online learning. We propose an efficient online algorithm and provide a formal analysis of its con- vergence properties. Moreover, first experimental results are presented, showing that our method performs well in practice.
Keywords: classification, learning theory, F-measure, structured out- put prediction
Copyright c 2015 by the paper’s authors. Copying permitted only for private and academic purposes. In: R. Bergmann, S. G¨org, G. M¨uller (Eds.): Proceedings of the LWA 2015 Workshops: KDML, FGWM, IR, and FGDB. Trier, Germany, 7.-9.
October 2015, published at http://ceur-ws.org
4 This work is accepted at the Twenty-ninth Annual Conference on Neural Information Processing Systems (NIPS).
106