The Geometry of ROC Space: Understanding Machine Learning Metrics through ROC Isometrics

By Peter A. Flach. International Conference on Machine Learning, 2003

318 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/b8eb7da56dae58f77788c33a57b5b810ca930527

How researchers see this paper

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

Citations in the corpus

Learning Decision Trees for Unbalanced Data (2008, methodology): “[18]” Subsequently, Flach [9] connected the isometric plots to ROC analysis, demonstrating the effects of true and false positives on several common evaluation metrics: accuracy, precision, and f-measure.”
An experimental comparison of performance measures for classification (2009, methodology): “Flach (2003), Fuernkranz and Flach (2005) analyse several metrics (AUC, accuracy, Fmeasure) using the ROC space.”
Information, Divergence and Risk for Binary Experiments (2009, methodology): “In this paper we draw connections between Receiver Operating Characteristic (ROC) curves, (Fawcett, 2006, 2004; Flach, 2003; Flach and Wu, 2005; Maxion and Roberts, 2004) the Area Under ROC Curve (AUC), (Cortes and Mohri, 2004; Hand, 2008; Hand and Till, 2001; Hanley and McNeil, 1982) and Cost…”
Precision-Recall-Gain Curves: PR Analysis Done Right (2015, methodology): “In particular, test set accuracy assuming uniform misclassiﬁcation costs is represented by accuracy isometrics with slope ( 1 − π ) / π , where π is the proportion of positives [5].”
Cost curves: An improved method for visualizing classifier performance (2006, background): “5 The construction shown on the right, due to Peter Flach (2003, 2004), draws a horizontal line segment from the y -axis to the point where the iso-performance line intersects the diagonal line TP = 1 − FP .”
Hellinger distance decision trees are robust and skew-insensitive (2011, methodology): “We point the interested reader to the paper by Flach for a more elaborate analysis of class skew using isometrics on these three metrics (Flach 2003).”
Measuring classifier performance: a coherent alternative to the area under the ROC curve (2009, methodology): “They include misclassification (or error) rate, the Kolmogorov-Smirnov (KS) statistic, likelihood ratios, the area under the ROC curve (or, equivalently, the Gini coefficient, defined below), pairs of measures such as specificity and sensitivity or precision and recall, measures of accuracy of proba”
Ethical Machine Learning in Healthcare (2020, methodology): “There are many commonly used performance metrics for model evaluation such as area under the receiver-operating characteristic (AUC), precision-recall curves (AUPRC), and calibration (113).”
Data mining in metric space: an empirical analysis of supervised learning performance criteria (2004, background): “The closest work to ours is by Flach [7].”