A Comparison of MCC and CEN Error Measures in Multi-Class Prediction

2010/08/17 by Giuseppe Jurman, Samantha Riccadonna, Cesare Furlanello
Computer Science · Mathematics · Psychology · #Anomaly Detection Techniques and Applications #Artificial intelligence #Class (philosophy) #Classifier (UML) #Computer science #Confusion #Correlation #Correlation coefficient #Data mining #Entropy (arrow of time) #Face and Expression Recognition #Generalization #Imbalanced Data Classification Techniques #Machine learning #Mathematics #Metric (unit) #Physics #Popularity #Psychology #stat.ML

paper · pdf · doi:10.1371/journal.pone.0041882

published as PLoS ONE 7(8): e41882 (2012)

arxiv created 2010/08/17 · openalex publication_date 2012/08/08 · arxiv updated 2012/08/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06

Abstract

We show that the Confusion Entropy, a measure of performance in multiclass problems has a strong (monotone) relation with the multiclass generalization of a classical metric, the Matthews Correlation Coefficient. Analytical results are provided for the limit cases of general no-information (n-face dice rolling) of the binary classification. Computational evidence supports the claim in the general case.

Citations

• Comparing Classification Results between N-ary and Binary Problems
• A Simple Generalisation of the Area Under the ROC Curve for Multiple Class Classification Problems
• A Mathematical Theory of Communication
• Using AUC and accuracy in evaluating learning algorithms
• Comparison of the predicted and observed secondary structure of T4 phage lysozyme
• The use of the area under the ROC curve in the evaluation of machine learning algorithms
• The meaning and use of the area under a receiver operating characteristic (ROC) curve.
• A systematic analysis of performance measures for classification tasks