I'm trying to build a predictive model, but I haven't found a method that consistently delivers high performance.
Is it acceptable to use an # Optimize classification threshold 0.996 ?
$\begingroup$ $\endgroup$
3 
- $\begingroup$ Questions in the Stack Exchange network work best when they focus on one question at a time. I've edited your post to just the first question. $\endgroup$Ben Reiniger– Ben Reiniger ♦2025-02-23 20:28:04 +00:00Commented Feb 23 at 20:28
- $\begingroup$ The second question of your original post has been asked and answered before, e.g. datascience.stackexchange.com/q/15630/55122 $\endgroup$Ben Reiniger– Ben Reiniger ♦2025-02-23 20:28:36 +00:00Commented Feb 23 at 20:28
- $\begingroup$ For the remaining question, some more details could be helpful. What's the level of class imbalance, what metrics are of interest for your problem (what is your problem?), and what are those metrics at the extreme threshold vs. some others? $\endgroup$Ben Reiniger– Ben Reiniger ♦2025-02-23 20:30:57 +00:00Commented Feb 23 at 20:30
Add a comment |