2. uber/orbit can receive arbitrary sparse training data (e.g.

train_ds=[2010, 2011, 2014, 2016]

) I was trying to build an adapter for orbit to be integrated into the nixtla model ecosystem (implement custom adapter class with

fit, predict

). However, I can’t seem to find examples where

train_ds

is also accessible, not just

train_y

. It seems that the only other option would be to (1) fill in the missing values in

y

with NaNs, so that the time series would be uniformly spaced with given freq, say

freq='Y'

, or (2) enforce myself to only uniformly spaced data in nixtla libraries (why isn’t this enforced already? Am I missing something)

This seams like a question for @fede (nixtla) (they/them).

Thank you for a quick reply, looking forward to the answer!

hey @Elijas! Currently,

statsforecast

does not support access to the

train_ds

values. Since our models don’t need the

ds

information, we omitted that possibility (they receive the time series as a numpy array). Even if the time series is not uniformly spaced, StatsForecast passes the values to the models. We are planning a refactor to allow the models to access the full dataframe (including

ds

) instead of just the numpy array values. Please help us open an issue to prioritize this new feature. 🙂

Where you thinking to open a PR in stats, ml o neural @Elijas?

Thank you for such a quick and clear response, @fede (nixtla) (they/them)!

Currently,

statsforecast

does not support access to the

train_ds

values

Just to make sure,

mlforecast

and

neuralforecast

also doesn’t support

ds

?

Thanks a lot @Elijas.

yes,

mlforecast

and

neuralforecast

also do not support it at the moment.

mlforecast

produces features based on

ds,

but the fit method of the models don’t receive the complete

ds

values