From b940bbc8a273032ca01091709acc83170db358e9 Mon Sep 17 00:00:00 2001
From: Tony Bagnall <ajb@uea.ac.uk>
Date: Fri, 18 Oct 2024 09:54:55 +0100
Subject: [PATCH] rework base class

---
 tsml_eval/_wip/forecasting/base.py        | 53 ++++++++++++++++++-----
 tsml_eval/_wip/forecasting/window_base.py | 26 +++++++----
 2 files changed, 60 insertions(+), 19 deletions(-)

diff --git a/tsml_eval/_wip/forecasting/base.py b/tsml_eval/_wip/forecasting/base.py
index f90a01fd..26b83aea 100644
--- a/tsml_eval/_wip/forecasting/base.py
+++ b/tsml_eval/_wip/forecasting/base.py
@@ -38,40 +38,71 @@ class BaseForecaster(BaseSeriesEstimator, ABC):
         predict value $window+i$. If horizon is 4, forecaster will used points $i$
         to $window+i-1$ to predict value $window+i+3$. If None, the algorithm will
         internally determine what data to use to predict `horizon` steps ahead.
-
-    To Do
-    -----
-    axis
     """
 
-    def __init__(self, horizon=1, window=None):
+    # TODO: add any forecasting specific tags
+    _tags = {
+        "capability:univariate": True,
+        "capability:multivariate": False,
+        "capability:missing_values": False,
+        "X_inner_type": "np.ndarray",  # one of VALID_INNER_TYPES
+    }
+    def __init__(self, horizon=1, window=None, axis=1):
         self.horizon = horizon
         self.window = window
         self._is_fitted = False
-        super().__init__(axis=1)
+        super().__init__(axis)
 
     @abstractmethod
-    def fit(self, X):
+    def fit(self, y, X=None):
         """Fit forecaster to series X.
+        TODO: passing series as X makes sense in machine learning, but not in
+        forecasting
 
         Parameters
         -------
-        X : np.ndarray
-            A time Time series on which to learn a forecaster
+        y : np.ndarray
+            A time series on which to learn a forecaster to predict horizon ahead
+        X : np.ndarray, default =None
+            Optional exogenous time series data assumed to be aligned with y
+
         Returns
         -------
         self
-            Fitted estimator
+            Fitted BaseForecaster.
         """
         ...
 
     @abstractmethod
-    def predict(self, X):
+    def predict(self, y=None, X=None):
         """
 
+        Parameters
+        -------
+        Parameters
+        -------
+        y : np.ndarray, default = None
+            A time series to predict the next horizon value for. If None,
+            predict the next horizon value after series seen in fit.
+        X : np.ndarray, default =None
+            Optional exogenous time series data assumed to be aligned with y
+
         Returns
         -------
         float
             single prediction.
         """
         ...
+
+    @abstractmethod
+    def forecast(self, y, X=None):
+        """
+
+        basically fit_predict.
+
+        Returns
+        -------
+        np.ndarray
+            single prediction directly after the last point in X.
+        """
+        ...
\ No newline at end of file
diff --git a/tsml_eval/_wip/forecasting/window_base.py b/tsml_eval/_wip/forecasting/window_base.py
index ad996c81..46377b71 100644
--- a/tsml_eval/_wip/forecasting/window_base.py
+++ b/tsml_eval/_wip/forecasting/window_base.py
@@ -8,10 +8,11 @@ def __init__(self, window, horizon=1, regressor=LinearRegression()):
         self.regressor = regressor
         super().__init__(horizon, window)
 
-    def fit(self, X):
-        """Fit forecaster to y, optionally using exogenous data X.
+    def fit(self, y, X=None):
+        """Fit forecaster to time series.
 
-        Split y into windows of length window and train the forecaster on each window
+        Split X into windows of length window and train the forecaster on each window
+        to predict the horizon ahead.
 
         Parameters
         ----------
@@ -23,17 +24,26 @@ def fit(self, X):
             Fitted estimator
         """
         # Window data
-        X2=np.lib.stride_tricks.sliding_window_view(X, window_shape=self.window)
-        X2=X2[:-self.horizon]
+        y2=np.lib.stride_tricks.sliding_window_view(y, window_shape=self.window)
+        # Ignore the final horizon values: need to store these for pred with empty y
+        y2=y2[:-self.horizon]
         # Extract y
         y=X[self.window+self.horizon-1:]
         self.regressor.fit(X2,y)
         return self
 
-    def predict(self, X):
+    def predict(self, y=None, X=None):
         """Predict values for time series X.
 
         NOTE: will not be able to predict the first
         """
-        X2 = np.lib.stride_tricks.sliding_window_view(X, window_shape=self.window)
-        return self.regressor.predict(X2[self.horizon:])
+        # TODO deal with case y =None
+        return self.regressor.predict(y[-self.window:])
+
+    def forecast(self, y, X=None):
+        """Forecast values for time series X.
+
+        NOTE: deal with horizons
+        """
+        self.fit(y,X)
+        return self.predict(y[:self.window])
\ No newline at end of file