Timeline for Philosophical question on logistic regression: why isn't the optimal threshold value trained?
Current License: CC BY-SA 4.0
2 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 19, 2023 at 14:52 | comment | added | cbeleites | It's implicitly covered in the cost/payoff, but I'd it deserves to be mentioned explicitly that those also depend on the prevalence/incidence/relative frequency of the classes under application conditions. At least in my experience, these are rarely the same as the total number of cases in the (true) classes in the confusion table/test data (and even rarer in training data). | |
| Apr 25, 2019 at 16:08 | history | answered | Stephan Kolassa | CC BY-SA 4.0 |