AUC: A Better Measure than Accuracy in Comparing Learning Algorithms

By Charles X. F. Ling, Jin Huang, Harry Zhang. Canadian Conference on AI, 2003

491 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

New methods for separating causes from effects in genomics data (2012, background): “The second metric is area under ROC curve (AUC), which is known to be more discriminative than accuracy because it takes into account the confidence of orientation decisions [30,31].”
An up-to-date comparison of state-of-the-art classification algorithms (2017, background): “This can be attributed to the fact that ACC ignores probability estimations of classiﬁcation in favour of class labels ( Ling, Huang, & Zhang, 2003 ), that is, the percentage of the correctly predicted instances for all classes is calculated collectively.”
Time-aware recommender systems: a comprehensive survey and analysis of existing evaluation protocols (2013, methodology): “…the information retrieval field, such as precision, recall, and normalized discounted cumulative gain (Baeza-Yates and Ribeiro-Neto 1999), and metrics used in the machine learning field, such as the receiver operating characteristic (ROC) curve, and the area under the ROC curve (Ling et al. 2003).”
PRIE: a system for generating rulelists to maximize ROC performance (2008, background): “Others have pointed out that accuracy is a poor metric to optimize (Provost et al. 1998; Ling et al. 2003), so accuracy-maximizing methods may be a poor choice for producing scoring classifiers—that is, classifiers with good ROC performance.”