[BUG-FIX] RegressionEnsembleModel : Error Furture covariates/Past Covariates… #1660

Rajesh4AI · 2023-03-21T16:54:01Z

… FutureCovariates | Suggested With Code Fix

[BUG-FIX] RegressionEnsembleModel : Error when trying to predict with FutureCovariates | Suggested With Code Fix #1621

Fixes #.

In regression_ensemble_model.py : While fitting the model the code did not fit with past/future covariates.
: Code has been modified to check if Covariates Exist and Model supports. If yes then fit with same.

Summary

Other Information

To reproduce the error try this:

from darts.models import RegressionEnsembleModel,AutoARIMA,Prophet
from darts.datasets import AirPassengersDataset
import numpy as np
series_air = AirPassengersDataset().load()
models = [Prophet(),AutoARIMA()]
ensemble_model = RegressionEnsembleModel(forecasting_models=models,regression_train_n_points=12)

Fit would run

ensemble_model.fit(series=series_air[:-36]
,future_covariates=series_air
)

Error: Predict would fail as Fit doesn't fit the models on future covariate due to bug

pred_air=ensemble_model.predict(n=36,series=series_air[:-36],future_covariates=series_air)

… FutureCovariates | Suggested With Code Fix

added space after comma

…/darts into fix_RegressionEnsemble

Rajesh4AI · 2023-03-23T05:25:06Z

@madtoinou Fixed the lint issue. Do let me know if there is anything else I need to check.

madtoinou

Thank you for contributing to darts, I did a little suggestion to simplify the modification.

Could you please also add a case in the unit test to cover this scenario (EnsembleModel containing LocalForecastingModels and using future covariates during training/prediction)?

madtoinou · 2023-03-23T08:41:45Z

darts/models/forecasting/regression_ensemble_model.py

+                elif (past_covariates is not None) and model.supports_past_covariates:
+                    model.fit(series=series, past_covariates=past_covariates)
+                else:
+                    model.fit(self.training_series)


if the if self.is_global_ensemble condition is not verified, it indicates that the EnsembleModel contains only LocalForecastingModels which do not support past covariates, you can remove this if-else case.

Also, you could probably also remove the check on past_covariate since a value of None will not disrupt the code-flow (as it's the default value in the fit() method).

A simpler approach would probably be to reuse the approach for the global_ensemble with something like:

if self.is_global_ensemble: kwargs = dict(series=series) if model.supports_past_covariates: kwargs["past_covariates"] = past_covariates else: kwargs = dict(series=self.training_series) if model.supports_future_covariates: kwargs["future_covariates"] = future_covariates model.fit(**kwargs)

Noice!! This is much clean... Will update it and write the test too, Thank you.

codecov-commenter · 2023-03-23T08:46:00Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 94.04%. Comparing base (ebb9eb6) to head (d888a37).
Report is 278 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1660      +/-   ##
==========================================
- Coverage   94.11%   94.04%   -0.07%     
==========================================
  Files         125      125              
  Lines       11447    11452       +5     
==========================================
- Hits        10773    10770       -3     
- Misses        674      682       +8

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

🚨 Try these New Features:

Flaky Tests Detection - Detect and resolve failed and flaky tests

for global ensemble it will check if supports past covariates. For both local and global model it will future covariates supports

Rajesh4AI · 2023-03-23T10:53:32Z

Working on the test case.
Moved kwargs above if condition.

            **kwargs = dict(series=series)**
            if self.is_global_ensemble:
                if model.supports_past_covariates:
                    kwargs["past_covariates"] = past_covariates
            if model.supports_future_covariates:
                kwargs["future_covariates"] = future_covariates
            model.fit(**kwargs)```

Tests with local models supports future covariates along with models which don't support. Prophet() is used to represent one which support future covariate

Added blank line before comment

Linted code with black

madtoinou · 2023-03-24T08:58:31Z

@Rajesh4AI, you can find a section at the bottom of the contribution guidelines explaining how to install the dev tools and automate the linting

fixed library import order

Rajesh4AI · 2023-03-24T11:47:18Z

@Rajesh4AI, you can find a section at the bottom of the contribution guidelines explaining how to install the dev tools and automate the lining

@madtoinou Sure, fixed the import sort manually in test_resgression_ensemble_model.py . For some reason auto linting is not working on my VScode (black - provider, isort = for sorting) . Will try from different system if it fails this time too.

Rajesh4AI · 2023-03-26T05:09:16Z

@madtoinou all linting issues have been fixed. Do let me know if i need to do anything else

madtoinou

Thank you for contributing to darts!

Suggested a minor modification, otherwise the code looks good 🚀

madtoinou · 2023-03-27T07:41:33Z

darts/tests/models/forecasting/test_regression_ensemble_model.py

+        ensemble_models = self.get_local_models()
+
+        # append model which supports future covariates
+        ensemble_models.append(Prophet())


There are discussion to make Prophet optional (imported from another library, installation is not necessarily straight-forward): it would probably to use a simpler model to test this feature such as LinearRegressionModel or RandomForest.

Makse sense.. will modify it

…/darts into fix_RegressionEnsemble

madtoinou

Thanks you for adapting the code so promptly!

dennisbader

Looks good, thanks :) I think we can even simplify a bit more :)

dennisbader · 2023-04-02T15:55:48Z

darts/models/forecasting/regression_ensemble_model.py

@@ -140,17 +140,13 @@ def fit(
        self.models = [model.untrained_model() for model in self.models]

        for model in self.models:
+            kwargs = dict(series=series)
            if self.is_global_ensemble:


couldn't we just fully ignore the self.is_global_ensemble and do the following for any model:

kwargs = dict(series=series) if model.supports_past_covariates: kwargs["past_covariates"] = past_covariates if model.supports_future_covariates: kwargs["future_covariates"] = future_covariates model.fit(**kwargs)

dennisbader

Thanks @Rajesh4AI , looks good 🚀

…ariates… (unit8co#1660) * [BUG-FIX] RegressionEnsembleModel : Error when trying to predict with FutureCovariates | Suggested With Code Fix * linted the code added space after comma * simplified future covariate check for global ensemble it will check if supports past covariates. For both local and global model it will future covariates supports * test RegressionEnsemble Model with future Covariates Tests with local models supports future covariates along with models which don't support. Prophet() is used to represent one which support future covariate * Fixed possible lint issue Added blank line before comment * Lint correction with Black Linted code with black * isort fix manual fixed library import order * Replace Prophet with altenatives * simplify covariate usage --------- Co-authored-by: madtoinou <32447896+madtoinou@users.noreply.github.com> Co-authored-by: dennisbader <dennis.bader@gmx.ch>

[BUG-FIX] RegressionEnsembleModel : Error when trying to predict with…

058c416

… FutureCovariates | Suggested With Code Fix

Rajesh4AI requested a review from dennisbader as a code owner March 21, 2023 16:54

Rajesh4AI changed the title ~~[BUG-FIX] RegressionEnsembleModel : Error when trying to predict with…~~ [BUG-FIX] RegressionEnsembleModel : Error Furture covariates/Past Covariates… Mar 22, 2023

madtoinou and others added 3 commits March 22, 2023 20:39

Merge branch 'master' into fix_RegressionEnsemble

aa27d52

linted the code

ea05a68

added space after comma

Merge branch 'fix_RegressionEnsemble' of https://github.com/Rajesh4AI…

23282fd

…/darts into fix_RegressionEnsemble

madtoinou reviewed Mar 23, 2023

View reviewed changes

simplified future covariate check

12ab02b

for global ensemble it will check if supports past covariates. For both local and global model it will future covariates supports

Rajesh4AI added 3 commits March 23, 2023 18:25

test RegressionEnsemble Model with future Covariates

92cacd3

Tests with local models supports future covariates along with models which don't support. Prophet() is used to represent one which support future covariate

Fixed possible lint issue

c5501ec

Added blank line before comment

Lint correction with Black

c90b337

Linted code with black

isort fix manual

2384d21

fixed library import order

Merge branch 'master' into fix_RegressionEnsemble

85e1f1e

Merge branch 'master' into fix_RegressionEnsemble

c8bebe5

madtoinou reviewed Mar 27, 2023

View reviewed changes

Rajesh4AI and others added 3 commits March 27, 2023 19:02

Replace Prophet with altenatives

1bd4339

Merge branch 'fix_RegressionEnsemble' of https://github.com/Rajesh4AI…

43bab6d

…/darts into fix_RegressionEnsemble

Merge branch 'master' into fix_RegressionEnsemble

f1e518a

madtoinou approved these changes Mar 27, 2023

View reviewed changes

dennisbader reviewed Apr 2, 2023

View reviewed changes

dennisbader added 2 commits April 10, 2023 18:13

Merge branch 'master' into fix_RegressionEnsemble

c0feee1

simplify covariate usage

48dc719

dennisbader approved these changes Apr 10, 2023

View reviewed changes

Merge branch 'master' into fix_RegressionEnsemble

d888a37

dennisbader merged commit 0401fe3 into unit8co:master Apr 10, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[BUG-FIX] RegressionEnsembleModel : Error Furture covariates/Past Covariates… #1660

[BUG-FIX] RegressionEnsembleModel : Error Furture covariates/Past Covariates… #1660

Rajesh4AI commented Mar 21, 2023 •

edited

Loading

Rajesh4AI commented Mar 23, 2023

madtoinou left a comment

madtoinou Mar 23, 2023

Rajesh4AI Mar 23, 2023

codecov-commenter commented Mar 23, 2023 •

edited by codecov bot

Loading

Rajesh4AI commented Mar 23, 2023 •

edited

Loading

madtoinou commented Mar 24, 2023

Rajesh4AI commented Mar 24, 2023

Rajesh4AI commented Mar 26, 2023 •

edited

Loading

madtoinou left a comment

madtoinou Mar 27, 2023

Rajesh4AI Mar 27, 2023

madtoinou left a comment

dennisbader left a comment

dennisbader Apr 2, 2023

dennisbader left a comment

[BUG-FIX] RegressionEnsembleModel : Error Furture covariates/Past Covariates… #1660

[BUG-FIX] RegressionEnsembleModel : Error Furture covariates/Past Covariates… #1660

Conversation

Rajesh4AI commented Mar 21, 2023 • edited Loading

Summary

Other Information

Fit would run

Error: Predict would fail as Fit doesn't fit the models on future covariate due to bug

Rajesh4AI commented Mar 23, 2023

madtoinou left a comment

Choose a reason for hiding this comment

madtoinou Mar 23, 2023

Choose a reason for hiding this comment

Rajesh4AI Mar 23, 2023

Choose a reason for hiding this comment

codecov-commenter commented Mar 23, 2023 • edited by codecov bot Loading

Codecov Report

Rajesh4AI commented Mar 23, 2023 • edited Loading

madtoinou commented Mar 24, 2023

Rajesh4AI commented Mar 24, 2023

Rajesh4AI commented Mar 26, 2023 • edited Loading

madtoinou left a comment

Choose a reason for hiding this comment

madtoinou Mar 27, 2023

Choose a reason for hiding this comment

Rajesh4AI Mar 27, 2023

Choose a reason for hiding this comment

madtoinou left a comment

Choose a reason for hiding this comment

dennisbader left a comment

Choose a reason for hiding this comment

dennisbader Apr 2, 2023

Choose a reason for hiding this comment

dennisbader left a comment

Choose a reason for hiding this comment

Rajesh4AI commented Mar 21, 2023 •

edited

Loading

codecov-commenter commented Mar 23, 2023 •

edited by codecov bot

Loading

Rajesh4AI commented Mar 23, 2023 •

edited

Loading

Rajesh4AI commented Mar 26, 2023 •

edited

Loading