Metrics for Multi-Class Classification: an Overview

From accuracy to Cohen-Kappa from Margherita Grandini et al. at CRIF

Original Blog Post: Metrics for Multi-Class Classification: an Overview
Authors: Margherita Grandini, Enrico Bagli, Giorgio Visani (from CRIF)
Published: 2020-08-14

My summary:

This paper is a must for anyone working on supervised, classification machine learning problems.

The paper goes into detail on classification metrics. The title specifies multi-class classification, but the text is relevant to binary classification as well.

Many data scientists are likely aware of the F1-score and the confusion matrix, but probably fewer are familiar with the Cohen-Kappa score. This paper goes through many classification metrics and discusses the pros and cons of each, with concrete examples.

Here is the link again to their paper for more details: Metrics for Multi-Class Classification: an Overview