Where academic tradition
meets the exciting future

You are here: TUCS > PUBLICATIONS > Publication Search > Efficient AUC Maximization wit...

Efficient AUC Maximization with Regularized Least-Squares

Tapio Pahikkala, Antti Airola, Hanna Suominen, Jorma Boberg, Tapio Salakoski, Efficient AUC Maximization with Regularized Least-Squares. In: Per Kreuger Anders Holst, Peter Funk (Eds.), Proceedings of the Tenth Scandinavian Conference on Artificial Intelligence, Frontiers in Artificial Intelligence and Applications 173, 12–19, IOS Press, 2008.

Abstract:

Area under the receiver operating characteristics curve (AUC) is a popular measure for evaluating the quality of binary classifiers, and intuitively, machine learning algorithms that maximize an approximation of AUC should have a good AUC performance when classifying new examples. However, designing such algorithms in the framework of kernel methods has proven to be challenging. In this paper, we address AUC maximization with the regularized least-squares (RLS) algorithm also known as the least-squares support vector machine. First, we introduce RLS-type binary classifier that maximizes an approximation of AUC and has a closed-form solution. Second, we show that this AUC-RLS algorithm is computationally as efficient as the standard RLS algorithm that maximizes an approximation of the accuracy. Third, we compare the performance of these two algorithms in the task of assigning topic labels for newswire articles in terms of AUC. Our algorithm outperforms the standard RLS in every classification experiment conducted. The performance gains are most substantial when the distribution of the class labels is unbalanced. In conclusion, modifying the RLS algorithm to maximize the approximation of AUC does not increase the computational complexity, and this alteration enhances the quality of the classifier.

Files:

Abstract in PDF-format

BibTeX entry:

@INPROCEEDINGS{inpPaAiSuBoSa08a,
  title = {Efficient AUC Maximization with Regularized Least-Squares},
  booktitle = {Proceedings of the Tenth Scandinavian Conference on Artificial Intelligence},
  author = {Pahikkala, Tapio and Airola, Antti and Suominen, Hanna and Boberg, Jorma and Salakoski, Tapio},
  volume = {173},
  series = {Frontiers in Artificial Intelligence and Applications},
  editor = {Anders Holst, Per Kreuger and Peter Funk},
  publisher = {IOS  Press},
  pages = {12–19},
  year = {2008},
  keywords = {Area under curve, Kernel methods, Regularized least-squares},
}

Belongs to TUCS Research Unit(s): Turku BioNLP Group

Publication Forum rating of this publication: level 1

Edit publication