add forecasting pipeline

feature_engine/timeseries/forecasting/pipeline.py

@@ -0,0 +1,116 @@
+import numpy as np
+import pandas as pd
+from sklearn.pipeline import Pipeline
+from sklearn.utils.validation import check_is_fitted
+from feature_engine.dataframe_checks import check_X,check_X_y
+
+
+class ForecastingPipeline(Pipeline):
+    """
+    TBD
+    """
+
+    def __init__(
+            self,
+            steps,
+            forecasting_horizon: int = None,
+            period: int = 1,
+            freq: str = None,
+            memory:bool=None,
+            verbose:bool=False,
+    ) -> None:
+
+        if not isinstance(forecasting_horizon, int):
+            raise ValueError(f"forecasting_horizon must be an integer. Got {forecasting_horizon} instead.")
+
+        if not isinstance(period, int):
+            raise ValueError(f"period must be an integer. Got {period} instead.")
+
+        if not isinstance(freq, str) or not freq in ["Y", "M", "D", "h", "m", "s"]:
+            raise ValueError(
+                "freq takes strings 'Y', 'M','D', 'h', 'm', 's'. "
+                f"Got {freq} instead."
+            )
+
+        self.steps = steps
+        self.forecasting_horizon = forecasting_horizon
+        self.period = period
+        self.freq = freq
+        self.memory = memory
+        self.verbose = verbose
+
+    def _get_steps(self):
+        _, estimators = zip(*self.steps)
+        return estimators
+
+    def _horizon(self, X):
+
+        forecast_start = X.index[-1] + np.timedelta64(self.period, self.freq)
+        forecast_end = forecast_start + np.timedelta64(self.forecasting_horizon, self.freq)
+
+        horizon = pd.date_range(
+            start=forecast_start,
+            end=forecast_end,
+            freq=self.freq,
+        )
+
+        return horizon
+
+    def fit(self, X, y):
+        Xt, y = check_X_y(X, y)
+
+        # Create features
+        for _, _, transform in self._iter():
+            if hasattr(transform, "transform"):
+                Xt = transform.fit_transform(Xt)
+
+            else:
+                # Check that input feature was removed
+                if any([f for f in X.columns if f in Xt.columns]):
+                    raise ValueError(
+                        "The predictive features contain the raw time series. "
+                    )
+                # Check if observations where removed after
+                # creating features, and if so, adjust y.
+                if len(Xt)!= len(y):
+                    y.loc[Xt.index]
+
+                # Fit regression model
+                transform.fit(Xt, y)
+
+        return self
+
+    def transform(self, X):
+
+        Xt = X.copy()
+        for _, _, transform in self._iter():
+            # need this tweak in case last class is not a
+            # transformer
+            if hasattr(transform, "transform"):
+                Xt = transform.transform(Xt)
+        return Xt
+
+    def forecast(self, X):
+
+        # check method fit has been called
+        check_is_fitted(self)
+
+        X = check_X(X)
+
+        horizon = self._horizon(X)
+
+        X = pd.concat([X, pd.DataFrame(index=horizon, columns=X.columns)], axis=0)
+
+        for step in horizon:
+
+            input_data = X[X.index < step]
+
+            X.loc[step] = self.predict(input_data)[-1]
+
+        return X.loc[horizon]
+
+    def get_feature_names_out(self, input_features=None):
+        if hasattr(self._get_steps()[-1], "transform"):
+            return self._get_steps()[-1].get_feature_names_out(input_features)
+        else:
+            return self._get_steps()[-2].get_feature_names_out(input_features)