Autoets forecast Feature suggestions

[vatsmanish]

Description

Hello Team ,
I would like to raise one of feature request in the AutoETS when we use . fit method we get forecast. but as we know we are passing multiple combinantion of model(e.g-(A,A,N),(M,A,N),(M,N,N)).

Now in the .fit method i have not specify any specific model , how would i see which of the above mentioned model has been selected??

Example:- I have passed 10 unique id in my model how would i figure out which unique id has selected what type of model has been fitted?

Use case

This feature will helps to enhance the model Interpretability and business to interpret the forecast. From this feature we can explain for different unique id which model has been selected and what will be there significance importance of this model?

[jmoralez]

Hey. When you call the fit method the models are stored in the fitted_ attribute, which is an array of shape (n_series, n_models). So for example you can get the model that was trained with the following:

from statsforecast import StatsForecast
from statsforecast.models import AutoETS
from statsforecast.utils import generate_series

series = generate_series(5, freq='M')
sf = StatsForecast(
    models=[AutoETS(season_length=12)],
    freq='M',
)
sf.fit(series)
fitted_ets = sf.fitted_[:, 0]  # first column here since it's the first model in the models list
model_by_id = {uid: m.model_['method'] for uid, m in zip(sf.uids, fitted_ets)}

[github-actions]

This issue has been automatically closed because it has been awaiting a response for too long. When you have time to to work with the maintainers to resolve this issue, please post a new comment and it will be re-opened. If the issue has been locked for editing by the time you return to it, please open a new issue and reference this one.