π οΈ Feature Engineering for Chronos-2 #336
abdulfatir
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Chronos-2 natively supports dynamic covariates. This opens up myriad possibilities for feature engineering to improve forecast accuracy. For the fev-bench results in the technical report, we just used the default covariates available for each task. However, better feature engineering could lead to further boost in the overall accuracy. I am starting this thread for discussion on feature engineering for Chronos-2. Share how your improved accuracy with better feature engineering or ask related questions.
Beta Was this translation helpful? Give feedback.
All reactions