The MCC-F1 curve: a performance evaluation technique for binary classification

Chang Cao, Davide Chicco, Michael M. Hoffman
6/17/2020
stat.MLcs.LG

Official code: hoffmangroup/mccf1 (4 stars, last commit 2020-05-20).

Abstract

Many fields use the ROC curve and the PR curve as standard evaluations of binary classification methods. Analysis of ROC and PR, however, often gives misleading and inflated performance evaluations, especially with an imbalanced ground truth. Here, we demonstrate the problems with ROC and PR analysis through simulations, and propose the MCC-F1 curve to address these drawbacks. The MCC-F1 curve combines two informative single-threshold metrics, MCC and the F1 score. The MCC-F1 curve more clearly differentiates good and bad classifiers, even with imbalanced ground truths. We also introduce the MCC-F1 metric, which provides a single value that integrates many aspects of classifier performance across the whole range of classification thresholds. Finally, we provide an R package that plots MCC-F1 curves and calculates related metrics.

View on arXivView PDF

Code Implementations(1)

hoffmangroup/mccf1Official100%
40RMay 4, 20206 years ago

Similar Papers

Overcoming Common Flaws in the Evaluation of Selective Classification Systems

83% match1 repo362024

Precision and Recall Reject Curves for Classification

82% match2023

Receiver operating characteristic (ROC) movies, universal ROC (UROC) curves, and coefficient of predictive ability (CPA)

82% match2019

On the Geometry of Receiver Operating Characteristic and Precision-Recall Curves

82% match2025

Which Metric Reflects the Spelling Rate Accuracy in Event-Related Potential-Based Brain-Computer Interfaces?

82% match2026

Area under the ROC Curve has the Most Consistent Evaluation for Binary Classification

80% match2024

Cite this paper

@article{cao2020mccf,
  title  = {The MCC-F1 curve: a performance evaluation technique for binary classification},
  author = {Chang Cao and Davide Chicco and Michael M. Hoffman},
  year   = {2020},
  eprint = {2006.11278},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2006.11278}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2006.11278.svg)](https://codeofpaper.com/paper/2006.11278)

Discussion