feat: add ARIMAPlus.predict_explain() to generate forecasts with explanation columns (#1177)

* feat: create arima_plus_predict_attribution method

* tmp: debug notes for time_series_arima_plus_model.predict_attribution

* update test_arima_plus_predict_explain_default test and create test_arima_plus_predict_explain_params test

* Merge branch 'ml-predict-explain' of github.com:googleapis/python-bigquery-dataframes into ml-predict-explain

* update  test_arima_plus_predict_explain_params test

* Revert "tmp: debug notes for time_series_arima_plus_model.predict_attribution"

This reverts commit f6dd455.

* format and lint

* Update bigframes/ml/forecasting.py

Co-authored-by: Tim Sweña (Swast) <[email protected]>

* update predict explain params test

* update test

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

* add unit test file - bare bones

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

* fixed tests

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

* lint

* lint

* fix test: float -> int

---------

Co-authored-by: Chelsea Lin <[email protected]>
Co-authored-by: Tim Sweña (Swast) <[email protected]>
Co-authored-by: Owl Bot <gcf-owl-bot[bot]@users.noreply.github.com>

Author: 4 people

bigframes/ml/core.py

@@ -172,6 +172,14 @@ def forecast(self, options: Mapping[str, int | float]) -> bpd.DataFrame:
         sql = self._model_manipulation_sql_generator.ml_forecast(struct_options=options)
         return self._session.read_gbq(sql, index_col="forecast_timestamp").reset_index()
 
+    def explain_forecast(self, options: Mapping[str, int | float]) -> bpd.DataFrame:
+        sql = self._model_manipulation_sql_generator.ml_explain_forecast(
+            struct_options=options
+        )
+        return self._session.read_gbq(
+            sql, index_col="time_series_timestamp"
+        ).reset_index()
+
     def evaluate(self, input_data: Optional[bpd.DataFrame] = None):
         sql = self._model_manipulation_sql_generator.ml_evaluate(
             input_data.sql if (input_data is not None) else None

bigframes/ml/forecasting.py

@@ -253,6 +253,43 @@ def predict(
             options={"horizon": horizon, "confidence_level": confidence_level}
         )
 
+    def predict_explain(
+        self, X=None, *, horizon: int = 3, confidence_level: float = 0.95
+    ) -> bpd.DataFrame:
+        """Explain Forecast time series at future horizon.
+
+        .. note::
+
+            Output matches that of the BigQuery ML.EXPLAIN_FORECAST function.
+            See: https://cloud.google.com/bigquery/docs/reference/standard-sql/bigqueryml-syntax-explain-forecast
+
+        Args:
+            X (default None):
+                ignored, to be compatible with other APIs.
+            horizon (int, default: 3):
+                an int value that specifies the number of time points to forecast.
+                The default value is 3, and the maximum value is 1000.
+            confidence_level (float, default 0.95):
+                A float value that specifies percentage of the future values that fall in the prediction interval.
+                The valid input range is [0.0, 1.0).
+
+        Returns:
+            bigframes.dataframe.DataFrame: The predicted DataFrames.
+        """
+        if horizon < 1:
+            raise ValueError(f"horizon must be at least 1, but is {horizon}.")
+        if confidence_level < 0.0 or confidence_level >= 1.0:
+            raise ValueError(
+                f"confidence_level must be [0.0, 1.0), but is {confidence_level}."
+            )
+
+        if not self._bqml_model:
+            raise RuntimeError("A model must be fitted before predict")
+
+        return self._bqml_model.explain_forecast(
+            options={"horizon": horizon, "confidence_level": confidence_level}
+        )
+
     @property
     def coef_(
         self,

bigframes/ml/sql.py

@@ -310,6 +310,14 @@ def ml_forecast(self, struct_options: Mapping[str, Union[int, float]]) -> str:
         return f"""SELECT * FROM ML.FORECAST(MODEL {self._model_ref_sql()},
   {struct_options_sql})"""
 
+    def ml_explain_forecast(
+        self, struct_options: Mapping[str, Union[int, float]]
+    ) -> str:
+        """Encode ML.EXPLAIN_FORECAST for BQML"""
+        struct_options_sql = self.struct_options(**struct_options)
+        return f"""SELECT * FROM ML.EXPLAIN_FORECAST(MODEL {self._model_ref_sql()},
+  {struct_options_sql})"""
+
     def ml_generate_text(
         self, source_sql: str, struct_options: Mapping[str, Union[int, float]]
     ) -> str:

tests/system/small/ml/test_forecasting.py

@@ -65,6 +65,42 @@ def test_arima_plus_predict_default(
     )
 
 
+def test_arima_plus_predict_explain_default(
+    time_series_arima_plus_model: forecasting.ARIMAPlus,
+):
+    utc = pytz.utc
+    predictions = time_series_arima_plus_model.predict_explain().to_pandas()
+    assert predictions.shape[0] == 369
+    predictions = predictions[
+        predictions["time_series_type"] == "forecast"
+    ].reset_index(drop=True)
+    assert predictions.shape[0] == 3
+    result = predictions[["time_series_timestamp", "time_series_data"]]
+    expected = pd.DataFrame(
+        {
+            "time_series_timestamp": [
+                datetime(2017, 8, 2, tzinfo=utc),
+                datetime(2017, 8, 3, tzinfo=utc),
+                datetime(2017, 8, 4, tzinfo=utc),
+            ],
+            "time_series_data": [2727.693349, 2595.290749, 2370.86767],
+        }
+    )
+    expected["time_series_data"] = expected["time_series_data"].astype(
+        pd.Float64Dtype()
+    )
+    expected["time_series_timestamp"] = expected["time_series_timestamp"].astype(
+        pd.ArrowDtype(pa.timestamp("us", tz="UTC"))
+    )
+
+    pd.testing.assert_frame_equal(
+        result,
+        expected,
+        rtol=0.1,
+        check_index_type=False,
+    )
+
+
 def test_arima_plus_predict_params(time_series_arima_plus_model: forecasting.ARIMAPlus):
     utc = pytz.utc
     predictions = time_series_arima_plus_model.predict(
@@ -96,6 +132,33 @@ def test_arima_plus_predict_params(time_series_arima_plus_model: forecasting.ARI
     )
 
 
+def test_arima_plus_predict_explain_params(
+    time_series_arima_plus_model: forecasting.ARIMAPlus,
+):
+    predictions = time_series_arima_plus_model.predict_explain(
+        horizon=4, confidence_level=0.9
+    ).to_pandas()
+    assert predictions.shape[0] >= 1
+    prediction_columns = set(predictions.columns)
+    expected_columns = {
+        "time_series_timestamp",
+        "time_series_type",
+        "time_series_data",
+        "time_series_adjusted_data",
+        "standard_error",
+        "confidence_level",
+        "prediction_interval_lower_bound",
+        "trend",
+        "seasonal_period_yearly",
+        "seasonal_period_quarterly",
+        "seasonal_period_monthly",
+        "seasonal_period_weekly",
+        "seasonal_period_daily",
+        "holiday_effect",
+    }
+    assert expected_columns <= prediction_columns
+
+
 def test_arima_plus_detect_anomalies(
     time_series_arima_plus_model: forecasting.ARIMAPlus, new_time_series_df
 ):

tests/unit/ml/test_forecasting.py

@@ -0,0 +1,58 @@
+# Copyright 2023 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import re
+
+import pytest
+
+from bigframes.ml import forecasting
+
+
+def test_predict_explain_low_confidence_level():
+    confidence_level = -0.5
+
+    model = forecasting.ARIMAPlus()
+
+    with pytest.raises(
+        ValueError,
+        match=re.escape(
+            f"confidence_level must be [0.0, 1.0), but is {confidence_level}."
+        ),
+    ):
+        model.predict_explain(horizon=4, confidence_level=confidence_level)
+
+
+def test_predict_high_explain_confidence_level():
+    confidence_level = 2.1
+
+    model = forecasting.ARIMAPlus()
+
+    with pytest.raises(
+        ValueError,
+        match=re.escape(
+            f"confidence_level must be [0.0, 1.0), but is {confidence_level}."
+        ),
+    ):
+        model.predict_explain(horizon=4, confidence_level=confidence_level)
+
+
+def test_predict_explain_low_horizon():
+    horizon = -1
+
+    model = forecasting.ARIMAPlus()
+
+    with pytest.raises(
+        ValueError, match=f"horizon must be at least 1, but is {horizon}."
+    ):
+        model.predict_explain(horizon=horizon, confidence_level=0.9)