Hacker News new | past | comments | ask | show | jobs | submit login
How and When to Use ROC Curves and Precision-Recall Curves for Classification (machinelearningmastery.com)
29 points by rch 9 months ago | hide | past | web | favorite | 6 comments



He calls it a "relative operating characteristic curve", then links to a Wikipedia article for the properly named "receiver operating characteristic curve". Maybe I'm wrong but I've never heard the former term used.


It's receiver. I believe it came about from operators looking radar receivers and fiddling with the gain. Low gain, nothing is a target, high gain everything is a target.


I like your interpretation as filter/signal logic.


it’s the first way we began thinking about ROC. My understanding agrees with the GP: this was a norbert wiener- era application of binary classifiers for “is there a submarine there, or not?” tasks.

ROC was next adopted by the psychophysics community. “Did you hear that tone, or not” is a pretty easy extension of the original task...and involved many of the same researchers after the war was over.


Fixed, thanks.


Me neither.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: