docs: add python code sample to multiple timeseries forecasting

[DevStephanie]

BEGIN_COMMIT_OVERRIDE
docs: Add python code sample for multiple forecasting time series (#531)
END_COMMIT_OVERRIDE

Thank you for opening a Pull Request! Before submitting your PR, there are a few things you can do to make sure it goes smoothly:

• Make sure to open an issue as a bug/issue before writing your code! That way we can discuss the change, evaluate designs, and agree on the general idea
• Ensure the tests and linter pass
• Code coverage does not decrease (if any source code was changed)
• Appropriate docs were updated (if necessary)

Fixes #<issue_number_goes_here> 🦕

[snippet-bot]

Here is the summary of changes.

You are about to add 2 region tags.

• samples/snippets/create_multiple_timeseries_forecasting_model.py:19, tag bigquery_dataframes_bqml_arima_multiple_step_2_visualize
• samples/snippets/create_multiple_timeseries_forecasting_model.py:72, tag bigquery_dataframes_bqml_arima_multiple_step_3_fit

---

This comment is generated by snippet-bot.
If you find problems with this result, please file an issue at:
https://github.com/googleapis/repo-automation-bots/issues.
To update this comment, add snippet-bot:force-run label or use the checkbox below:

• Refresh this comment

[DevStephanie]

Corrected region tags, will be updated on next PR!

[GarrettWu]

Already fixed

[GarrettWu]

If you don't need filters, no need to pass in.

[DevStephanie]

ok, will correct.

[GarrettWu]

Why use read_gbq_table instead of more general API read_gbq?

[GarrettWu]

to_frame() not needed.

[DevStephanie]

ok, will update that.

[DevStephanie]

Looks like we might need .to_frame() because without it, I see

AttributeError                            Traceback (most recent call last)
Cell In[12], line 18
     15 X = num_trips["date"]
     16 y = num_trips["num_trips"]
---> 18 model.fit(X, y)
     19 # The model.fit() call above created a temporary model.
     20 # Use the to_gbq() method to write to a permanent location.
     22 model.to_gbq(
     23     your_model_id,  # For example: "bqml_tutorial.sample_model",
     24     replace=True,
     25 )

File ~/python-bigquery-dataframes/bigframes/ml/base.py:163, in SupervisedTrainablePredictor.fit(self, X, y)
    158 def fit(
    159     self: _T,
    160     X: Union[bpd.DataFrame, bpd.Series],
    161     y: Union[bpd.DataFrame, bpd.Series],
    162 ) -> _T:
--> 163     return self._fit(X, y)

File ~/python-bigquery-dataframes/bigframes/core/log_adapter.py:44, in method_logger.<locals>.wrapper(*args, **kwargs)
     42 if api_method_name.startswith("__") or not api_method_name.startswith("_"):
     43     add_api_method(full_method_name)
---> 44 return method(*args, **kwargs)

File ~/python-bigquery-dataframes/bigframes/ml/forecasting.py:218, in ARIMAPlus._fit(self, X, y, transforms)
    197 def _fit(
    198     self,
    199     X: Union[bpd.DataFrame, bpd.Series],
    200     y: Union[bpd.DataFrame, bpd.Series],
    201     transforms: Optional[List[str]] = None,
    202 ):
    203     """Fit the model to training data.
    204 
    205     Args:
   (...)
    216         ARIMAPlus: Fitted estimator.
    217     """
--> 218     if X.columns.size != 1:
    219         raise ValueError(
    220             "Time series timestamp input X must only contain 1 column."
    221         )
    222     if y.columns.size != 1:

File ~/python-bigquery-dataframes/bigframes/series.py:1062, in Series.__getattr__(self, key)
   1053     raise AttributeError(
   1054         textwrap.dedent(
   1055             f"""
   (...)
   1059         )
   1060     )
   1061 else:
-> 1062     raise AttributeError(key)

AttributeError: columns

[GarrettWu]

Why code below are all commented out?

[tswast]

This is just reading a regular table. We don't need to explain wildcard tables.

Also, the default filter is no filters, so we don't need to pass that in.

Suggested change

	# either a SQL query or a table ID. Since this example selects from multiple
	# tables via a wildcard, use SQL to define this data. Watch issue
	# https://github.com/googleapis/python-bigquery-dataframes/issues/169
	# for updates to `read_gbq_table` to support wildcard tables.
	df = bpd.read_gbq_table("bigquery-public-data.new_york.citibike_trips", filters=[])
	# either a SQL query or a table ID.
	df = bpd.read_gbq_table("bigquery-public-data.new_york.citibike_trips")

[DevStephanie]

Ok, will edit the filter but removing from this section that doesn't have one and removing the read gbp table explanation.

[DevStephanie]

Resolved

[tswast]

As discussed in our 1:1, these region tags need to be globally unique, so let's borrow from the URL of the tutorial (https://cloud.google.com/bigquery/docs/arima-multiple-time-series-forecasting-tutorial) to construct these.

For example:

Suggested change

	# [START bigquery_dataframes_bqml_create_data__set]
	# [START bigquery_dataframes_bqml_arima_multiple_step_2_visualize]

[DevStephanie]

yes, region tags will follow URL.

[DevStephanie]

Corrected

[tswast]

These two region tags can be combined since they are both for step 2.

[DevStephanie]

ok, makes sense! is "/ " or "&" the preference?

[tswast]

Let's stay consistent with the SQL with regards to suggested model ID. See step 3: https://cloud.google.com/bigquery/docs/arima-multiple-time-series-forecasting-tutorial#arima-single-model

Suggested change

	your_model_id, # For example: "bqml_tutorial.sample_model",
	your_model_id, # For example: "bqml_tutorial.nyc_citibike_arima_model",

[DevStephanie]

OOH, good catch! Yes, will correct this.

[tswast]

Not sure what this comment is from. Let's remove it.

Suggested change

	# ARIMAPlus(auto_arima_max_order=5, data_frequency='AUTO_FREQUENCY',
	# max_time_series_length=3, min_time_series_length=20,
	# time_series_length_fraction=1.0, trend_smoothing_window_size=-1)

[tswast]

As discussed in our 1:1, EXPLAIN_FORECAST isn't implemented yet in bigframes. Please file an issue if you haven't already and remove this sample for now.

[DevStephanie]

Yes, will remove and file request.

[tswast]

Thanks! Will wait for e2e tests with code samples to pass before merging.

[tswast]

e2e tests failed with

FAILED tests/system/large/ml/test_core.py::test_bqml_e2e - AssertionError: Da...
FAILED tests/system/large/ml/test_ensemble.py::test_xgbregressor_default_params
FAILED tests/system/large/ml/test_pipeline.py::test_pipeline_random_forest_classifier_fit_score_predict
FAILED tests/system/large/ml/test_pipeline.py::test_pipeline_xgbregressor_fit_score_predict

which appear to be unrelated.