XGBoost Support for Forecast Operator

.github/workflows/run-forecast-unit-tests.yml

@@ -27,14 +27,16 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.9", "3.10", "3.11"]
+        python-version: ["3.10", "3.11"]
 
     steps:
       - uses: actions/checkout@v4
         with:
           fetch-depth: 0
           ref: ${{ github.event.pull_request.head.sha }}
 
+      - uses: ./.github/workflows/create-more-space
+        name: "Create more disk space"
 
       - uses: actions/setup-python@v5
         with:

.github/workflows/run-operators-unit-tests.yml

@@ -27,7 +27,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.9", "3.10", "3.11"]
+        python-version: ["3.10", "3.11"]
 
     steps:
       - uses: actions/checkout@v4

ads/opctl/operator/lowcode/common/utils.py

@@ -327,7 +327,7 @@ def get_frequency_of_datetime(dt_col: pd.Series, ignore_duplicates=True):
     str  Pandas Datetime Frequency
     """
     s = pd.Series(dt_col).drop_duplicates() if ignore_duplicates else dt_col
-    return pd.infer_freq(s)
+    return pd.infer_freq(s) or pd.infer_freq(s[-5:])
 
 
 def human_time_friendly(seconds):

ads/opctl/operator/lowcode/forecast/const.py

@@ -14,6 +14,7 @@ class SupportedModels(ExtendedEnum):
     Arima = "arima"
     NeuralProphet = "neuralprophet"
     LGBForecast = "lgbforecast"
+    XGBForecast = "xgbforecast"
     AutoMLX = "automlx"
     Theta = "theta"
     ETSForecaster = "ets"

ads/opctl/operator/lowcode/forecast/model/base_model.py

@@ -611,7 +611,7 @@ def get_path(filename):
                     results.set_global_explanations(global_expl_rounded)
                 else:
                     logger.warning(
-                        f"Attempted to generate global explanations for the {self.spec.global_explanation_filename} file, but an issue occured in formatting the explanations."
+                        f"Attempted to generate global explanations for the {self.spec.global_explanation_filename} file, but an issue occurred in formatting the explanations."
                     )
 
                 if not self.formatted_local_explanation.empty:
@@ -634,7 +634,7 @@ def get_path(filename):
                     results.set_local_explanations(local_expl_rounded)
                 else:
                     logger.warning(
-                        f"Attempted to generate local explanations for the {self.spec.local_explanation_filename} file, but an issue occured in formatting the explanations."
+                        f"Attempted to generate local explanations for the {self.spec.local_explanation_filename} file, but an issue occurred in formatting the explanations."
                     )
             except AttributeError as e:
                 logger.warning(

ads/opctl/operator/lowcode/forecast/model/factory.py

@@ -20,9 +20,10 @@
 from .autots import AutoTSOperatorModel
 from .base_model import ForecastOperatorBaseModel
 from .forecast_datasets import ForecastDatasets
-from .ml_forecast import MLForecastOperatorModel
+from .lgbforecast import LGBForecastOperatorModel
 from .neuralprophet import NeuralProphetOperatorModel
 from .prophet import ProphetOperatorModel
+from .xgbforecast import XGBForecastOperatorModel
 from .theta import ThetaOperatorModel
 from .ets import ETSOperatorModel
 
@@ -45,7 +46,8 @@ class ForecastOperatorModelFactory:
         SupportedModels.Prophet: ProphetOperatorModel,
         SupportedModels.Arima: ArimaOperatorModel,
         SupportedModels.NeuralProphet: NeuralProphetOperatorModel,
-        SupportedModels.LGBForecast: MLForecastOperatorModel,
+        SupportedModels.LGBForecast: LGBForecastOperatorModel,
+        SupportedModels.XGBForecast: XGBForecastOperatorModel,
         SupportedModels.AutoMLX: AutoMLXOperatorModel,
         SupportedModels.AutoTS: AutoTSOperatorModel,
         SupportedModels.Theta: ThetaOperatorModel,

ads/opctl/operator/lowcode/forecast/model/forecast_datasets.py

@@ -26,7 +26,6 @@ class HistoricalData(AbstractData):
     def __init__(self, spec, historical_data=None, subset=None):
         super().__init__(spec=spec, name="historical_data", data=historical_data, subset=subset)
         self.subset = subset
-        self.freq = None
 
     def _ingest_data(self, spec):
         try:

ads/opctl/operator/lowcode/forecast/model/lgbforecast.py

@@ -0,0 +1,179 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2024 Oracle and/or its affiliates.
+# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
+import logging
+import traceback
+
+import pandas as pd
+
+from ads.common.decorator import runtime_dependency
+from ads.opctl import logger
+from .forecast_datasets import ForecastDatasets, ForecastOutput
+from .ml_forecast import MLForecastBaseModel
+from ..const import ForecastOutputColumns, SupportedModels
+from ..operator_config import ForecastOperatorConfig
+
+
+class LGBForecastOperatorModel(MLForecastBaseModel):
+    """Class representing MLForecast operator model."""
+
+    def __init__(self, config: ForecastOperatorConfig, datasets: ForecastDatasets):
+        super().__init__(config=config, datasets=datasets)
+
+    def get_model_kwargs(self):
+        """
+        Returns the model parameters.
+        """
+        model_kwargs = self.spec.model_kwargs
+
+        upper_quantile = round(0.5 + self.spec.confidence_interval_width / 2, 2)
+        lower_quantile = round(0.5 - self.spec.confidence_interval_width / 2, 2)
+
+        model_kwargs["lower_quantile"] = lower_quantile
+        model_kwargs["upper_quantile"] = upper_quantile
+        return model_kwargs
+
+
+    def preprocess(self, df, series_id):
+        pass
+
+    @runtime_dependency(
+        module="mlforecast",
+        err_msg="MLForecast is not installed, please install it with 'pip install mlforecast'",
+    )
+    @runtime_dependency(
+        module="lightgbm",
+        err_msg="lightgbm is not installed, please install it with 'pip install lightgbm'",
+    )
+    def _train_model(self, data_train, data_test, model_kwargs):
+        import lightgbm as lgb
+        from mlforecast import MLForecast
+        try:
+
+            lgb_params = {
+                "verbosity": model_kwargs.get("verbosity", -1),
+                "num_leaves": model_kwargs.get("num_leaves", 512),
+            }
+
+            data_freq = self.datasets.get_datetime_frequency()
+
+            additional_data_params = self.set_model_config(data_freq, model_kwargs)
+
+            fcst = MLForecast(
+                models={
+                    "forecast": lgb.LGBMRegressor(**lgb_params),
+                    "upper": lgb.LGBMRegressor(
+                        **lgb_params,
+                        objective="quantile",
+                        alpha=model_kwargs["upper_quantile"],
+                    ),
+                    "lower": lgb.LGBMRegressor(
+                        **lgb_params,
+                        objective="quantile",
+                        alpha=model_kwargs["lower_quantile"],
+                    ),
+                },
+                freq=data_freq,
+                date_features=['year', 'month', 'day', 'dayofweek', 'dayofyear'],
+                **additional_data_params,
+            )
+
+            num_models = model_kwargs.get("recursive_models", False)
+
+            self.model_columns = [
+                ForecastOutputColumns.SERIES
+            ] + data_train.select_dtypes(exclude=["object"]).columns.to_list()
+            fcst.fit(
+                data_train[self.model_columns],
+                static_features=model_kwargs.get("static_features", []),
+                id_col=ForecastOutputColumns.SERIES,
+                time_col=self.date_col,
+                target_col=self.spec.target_column,
+                fitted=True,
+                max_horizon=None if num_models is False else self.spec.horizon,
+            )
+
+            self.outputs = fcst.predict(
+                h=self.spec.horizon,
+                X_df=pd.concat(
+                    [
+                        data_test[self.model_columns],
+                        fcst.get_missing_future(
+                            h=self.spec.horizon, X_df=data_test[self.model_columns]
+                        ),
+                    ],
+                    axis=0,
+                    ignore_index=True,
+                ).fillna(0),
+            )
+            self.fitted_values = fcst.forecast_fitted_values()
+            for s_id in self.datasets.list_series_ids():
+                self.forecast_output.init_series_output(
+                    series_id=s_id,
+                    data_at_series=self.datasets.get_data_at_series(s_id),
+                )
+
+                self.forecast_output.populate_series_output(
+                    series_id=s_id,
+                    fit_val=self.fitted_values[
+                        self.fitted_values[ForecastOutputColumns.SERIES] == s_id
+                    ].forecast.values,
+                    forecast_val=self.outputs[
+                        self.outputs[ForecastOutputColumns.SERIES] == s_id
+                    ].forecast.values,
+                    upper_bound=self.outputs[
+                        self.outputs[ForecastOutputColumns.SERIES] == s_id
+                    ].upper.values,
+                    lower_bound=self.outputs[
+                        self.outputs[ForecastOutputColumns.SERIES] == s_id
+                    ].lower.values,
+                )
+
+                self.model_parameters[s_id] = {
+                    "framework": SupportedModels.LGBForecast,
+                    **lgb_params,
+                    **fcst.models_['forecast'].get_params(),
+                }
+
+            logger.debug("===========Done===========")
+
+        except Exception as e:
+            self.errors_dict[self.spec.model] = {
+                "model_name": self.spec.model,
+                "error": str(e),
+                "error_trace": traceback.format_exc(),
+            }
+            logger.warning(f"Encountered Error: {e}. Skipping.")
+            logger.warning(traceback.format_exc())
+            raise e
+
+
+    def _generate_report(self):
+        """
+        Generates the report for the model
+        """
+        import report_creator as rc
+
+        logging.getLogger("report_creator").setLevel(logging.WARNING)
+
+        # Section 2: LGBForecast Model Parameters
+        sec2_text = rc.Block(
+            rc.Heading("LGBForecast Model Parameters", level=2),
+            rc.Text("These are the parameters used for the LGBForecast model."),
+        )
+
+        k, v = next(iter(self.model_parameters.items()))
+        sec_2 = rc.Html(
+            pd.DataFrame(list(v.items())).to_html(index=False, header=False),
+        )
+
+        all_sections = [sec2_text, sec_2]
+        model_description = rc.Text(
+            "LGBForecast uses mlforecast framework to perform time series forecasting using machine learning models"
+            "with the option to scale to massive amounts of data using remote clusters."
+            "Fastest implementations of feature engineering for time series forecasting in Python."
+            "Support for exogenous variables and static covariates."
+        )
+
+        return model_description, all_sections