Revert "User Specified Date Feature"

[ParthivNaresh]

Fixes #2215
Reverts #2155

Reverting until we can figure out what's going on with pmdarima during installation

[dsherry]

Thanks for keeping main green! Sorry you ran into trouble with conda. We can fix it 😁

[freddyaboulton]

@ParthivNaresh @dsherry What happened?

[ParthivNaresh]

@freddyaboulton Filed #2215

[freddyaboulton]

@ParthivNaresh Why did the "Date Feature" PR add pmdarima as a dependency if the Arima implementation still used statsmodels? I think that's what's causing the conda test failure. pmdarima is listed in our requirements.txt but not in conda recipe (I had suggested we keep pmdarima off the conda recipe since I thought it wasn't needed)

[ParthivNaresh]

@freddyaboulton We moved ARIMA from the statsmodel implementation to sktime which has pmdarima as a soft dependency. So I guess this means we don't need statsmodel?

[freddyaboulton]

Ah I read the diff wrong! The original date feature pr changed the underlying library used by Arima.

Maybe what we should do then is add pmdarima to the conda recipe? Might need to remove statsmodels from our recipe though. I see pmdarima has a max statsmodels version that's lower than our min.

[codecov]

Codecov Report

Merging #2214 (eac0de7) into main (1ed595e) will decrease coverage by 0.1%.
The diff coverage is 100.0%.

@@            Coverage Diff            @@
##             main    #2214     +/-   ##
=========================================
- Coverage   100.0%   100.0%   -0.0%     
=========================================
  Files         296      296             
  Lines       24871    24712    -159     
=========================================
- Hits        24853    24694    -159     
  Misses         18       18             

Impacted Files	Coverage Δ
evalml/automl/utils.py	100.0% <ø> (ø)
evalml/preprocessing/utils.py	100.0% <ø> (ø)
.../tests/component_tests/test_datetime_featurizer.py	100.0% <ø> (ø)
evalml/tests/conftest.py	100.0% <ø> (ø)
...s/prediction_explanations_tests/test_explainers.py	100.0% <ø> (ø)
evalml/utils/gen_utils.py	99.6% <ø> (ø)
evalml/automl/automl_search.py	100.0% <100.0%> (ø)
...omponents/estimators/regressors/arima_regressor.py	100.0% <100.0%> (ø)
.../transformers/preprocessing/datetime_featurizer.py	100.0% <100.0%> (ø)
...rmers/preprocessing/delayed_feature_transformer.py	100.0% <100.0%> (ø)
... and 21 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1ed595e...eac0de7. Read the comment docs.

[ParthivNaresh]

@freddyaboulton If statsmodel isn’t required by anything else then I agree, we can take statsmodel out of the recipe. This might also fix the errors I was seeing when I added pmdarima to the recipe. There were a large number of dependency version mismatch issues which may have stemmed from its conflict with statsmodel.

[freddyaboulton]

@ParthivNaresh Yea I don't think statsmodels is used by anything else. I think it's worth seeing if we can patch up the recipe before merging this pr?

[ParthivNaresh]

@freddyaboulton Sorry I didn’t see your message before merging (dang refreshing). But I’ll take a look at the recipe next for sure