Recall vs Precision

Precision is obvious:
    "What noise came with this?"
Recall is subtle:
    "How much did we miss?"

In brief and for practical purposes, look at what we have versus what we omitted:

It's often easy to measure Precision, since we've got the sample and can study it. It's often hard to measure Recall, since it requires knowing (or at least estimating) what was missing from the sample.

It's often easy to get perfect Precision: just find one good thing and declare victory with that tiny sample! It's often easy to get total Recall: grab everything, including junk, and then nothing good is missed!

There are a host of other ratios to take when sampling the universe of Good Stuff and Bad Stuff. For much too much terminology see Wikipedia on precision and recall, sensitivity and specificity, false positives and false negatives, Type I and Type II errors, and the aptly named Confusion Matrix.

Or just remember to look at both Knowns and Unknowns — and don't be too sure about the difference!

(cf Knowns and Unknowns (1999-07-11), Unknown Knowns (2008-08-29), In the Presence of Oxygen (2017-06-15), ...) - ^z - 2020-01-08