# What is the name for the complement of accuracy?

I have a metric that is defined as $$1 - Accuracy$$ and I need a name for it. Is there a scientific name for the complement of accuracy?

• Seems as if there may be more than one definition for accuracy in common usage. Which do you mean? – Avraham Dec 12 '18 at 19:48
• I mean specifically accuracy. I have a metric that is defined in academics as blah_blah_accuracy and I can only compute 1-X of that metric. So, I was curious for the definition of the inverse of accuracy to call my metric blah_blah_(inverse of accuracy) – nikolaevra Dec 12 '18 at 23:04
• I don't mean to be a stickler, but what you're describing here isn't an 'inverse', it's a 'complement' (or, a particular type of complement if you're going down proper fuzzy theory, albeit 1-a is the commonest version used and the one typically implied unless explicitly specified otherwise). In fact, 'the complement of the accuracy' is a perfectly valid description for it and could be notated as $A^c$ (if accuracy is notated as $A$). – Tasos Papastylianou Dec 12 '18 at 23:29
• @TasosPapastylianou is right. You are looking for the complement, so long as $A$, and thus $A^c$ or $\bar{A}$ is restricted to $[0, 1]$. – Avraham Dec 12 '18 at 23:52
• Right, I will update the question – nikolaevra Dec 13 '18 at 3:29

I've seen people use $$\text{error rate} = 1 - \text{accuracy}$$, on the premise that accuracy is the proportion of samples classified correctly, so the error rate is the proportion of samples classified incorrectly.