Unachievable Region in Precision-Recall Space and Its Effect on Empirical Evaluation

TL;DR

This paper identifies an unachievable region in precision-recall space that depends solely on class skew, impacting how machine learning models are evaluated using PR curves.

Contribution

It precisely characterizes the unachievable region in PR space and discusses its implications for empirical evaluation methodology.

Findings

Unachievable region size depends only on class skew

PR curves vary with class skew

Implications for evaluation methodology

Abstract

Precision-recall (PR) curves and the areas under them are widely used to summarize machine learning results, especially for data sets exhibiting class skew. They are often used analogously to ROC curves and the area under ROC curves. It is known that PR curves vary as class skew changes. What was not recognized before this paper is that there is a region of PR space that is completely unachievable, and the size of this region depends only on the skew. This paper precisely characterizes the size of that region and discusses its implications for empirical evaluation methodology in machine learning.