0

When using least square for classification of two classes labeled as 0 and 1, the value 0.5 is a common choice for the threshold.

Intuitively, it makes sense, but is it always optimal? In which sense?

toliveira
  • 287
  • 1
  • 9
  • 2
    You might want to look into the “proper scoring rules” that get talked about on here sometimes. I also like Frank Harrell’s blog: https://www.fharrell.com/post/class-damage/ https://www.fharrell.com/post/classification/ – Dave Mar 09 '21 at 03:19
  • 2
    Some relevant discussions: https://stats.stackexchange.com/questions/312780/why-is-accuracy-not-the-best-measure-for-assessing-classification-models and https://stats.stackexchange.com/questions/105501/understanding-roc-curve/105577#105577 – Sycorax Mar 09 '21 at 03:19
  • 2
    [Here is a previous thread on thresholds.](https://stats.stackexchange.com/q/312119/1352) – Stephan Kolassa Mar 09 '21 at 06:35
  • See also: https://stats.stackexchange.com/questions/123124/how-to-determine-the-optimal-threshold-for-a-classifier-and-generate-roc-curve – Sycorax Mar 09 '21 at 17:06

0 Answers0