Commentary: Dabblers: Beware of hidden dangers in machine-learning comparisons.

Ishwaran, Hemant; Blackstone, Eugene H · J Thorac Cardiovasc Surg · 2022

editorial · Level V

Where this comes from

Abstract

Machine learning is not for dabblers. An underappreciated problem is imbalance between number of events and non-events, for which traditional C-statistics is an inappropriate evaluation metric. Characters + spaces: 193/200 Misleading C-statistics (AUC) versus precision-recall (PC) and G mean for imbalanced data. Characters + spaces: 90/90

Medical subject headings