correct approach in forecasting ranges in multiclassification #2432
Unanswered
algomaschine
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Gents, I'm predicting a float column that has distribution pretty much like below.
So there are different approaches that I'm testing (I'm really interested mostly in the direction and magnitude, not precise values).
split the target into classes like {-50,-40,30,20,10, 10,20,30,40,50} - and make 10 binary columns, do multiclassification within 1 model.
just assign the categories within one column {-50,-40,30,20,10, 0, 10,20,30,40,50} and forecast the category using 1 column.
make 1 distinct binary model per each category (as in classic methods before multiclassification existed).
Now I'm trying different variations and also different splittage of classes. However, is there a catboost official guideline which method is the best with this package? Been watching videos, reading some docs, but haven't really found an answer...
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions