Improve description for ARIMA parameters (p, q, seasonal_orders and trend)

[MarcBresson]

Summary

It provides more context to some features that are available in statsmodels but were not mentionned in Darts documentation.

Other Information

See statsmodels/statsmodels#9065 that

[MarcBresson]

Black has nothing to reformat but gradle still fails.

[dennisbader]

Hi @MarcBresson, simply run black like below (we don't use line length 120)

black darts/models/forecasting.arima.py

[dennisbader]

@MarcBresson , now isort is complaining. I suggest installing the pre-commit hook for automatic lint/reformatting as described here

[MarcBresson]

I'm sorry for all the troubles. I had issues with flake on the pre-commit hook, so I took the liberty to update it to the last version.

I forced push to clean the commit history and regroup all the isort and black reformatting under the same commit

[MarcBresson]

All the tests passed, I rebased the branch so its ready to be merged

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (8073de4) 93.88% compared to head (c5fa3fb) 93.87%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2142      +/-   ##
==========================================
- Coverage   93.88%   93.87%   -0.02%     
==========================================
  Files         135      135              
  Lines       13429    13420       -9     
==========================================
- Hits        12608    12598      -10     
- Misses        821      822       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[MarcBresson]

Should I update the test with

def test_historical_forecasts_transferrable_future_cov_local_models(self):
    model = ARIMA(p=np.array([1, 2, 3, 4, 6]), q=(2, 3))
    ...

?

[madtoinou]

Instead of modifying the current test, I would rather add new one so that the difference between the two set of parameters can be highlighted/verified.

[dennisbader]

Thanks a lot for this nice contribution @MarcBresson, I didn't even know that you could use a sequence of lags for more fine grained control over the parameters. That's great!

I added some suggestions that are just trying to rephrase some bits here and there to follow more our style from other models.

After that it's ready to be merged 🚀

[dennisbader]

let's add this to the CHANGELOG.md under
UNRELEASED -> For developers of the library

[MarcBresson]

Should I update other pre-commit hooks ? That way we are up to date and developers will not have to pre-commit install anytime soon.

[MarcBresson]

I updated all pre-commit hooks and change the argument for pyupgrade. See changelog (or commit description) for more info

[dennisbader]

Hi @MarcBresson, thanks for the changes. Everything loks fine now.

As a last point, could you move the changes to the pre commit hooks to a separate PR? Like that we keep it clearly separated.

After that we can merge :) 🚀

[dennisbader]

Thanks a lot @MarcBresson for this contribution, now everything looks perfect :) 🚀

Sorry if the journey was a bit long for this. I'll merge now.