[ENH] fit_predict for FreshPRINCE and RotationForest (#1456)

* fp and rotf fit_predict

* fix

Author: MatthewMiddlehurst

aeon/classification/feature_based/_fresh_prince.py

@@ -9,7 +9,6 @@
 
 
 import numpy as np
-from sklearn.utils import check_random_state
 
 from aeon.classification.base import BaseClassifier
 from aeon.classification.sklearn import RotationForestClassifier
@@ -117,8 +116,8 @@ def _fit(self, X, y):
         Changes state by creating a fitted model that updates attributes
         ending in "_" and sets is_fitted flag to True.
         """
-        self._fit_fresh_prince(X, y)
-
+        X_t = self._fit_fp_shared(X, y)
+        self._rotf.fit(X_t, y)
         return self
 
     def _predict(self, X) -> np.ndarray:
@@ -155,24 +154,18 @@ def _predict_proba(self, X) -> np.ndarray:
         return self._rotf.predict_proba(self._tsfresh.transform(X))
 
     def _fit_predict(self, X, y) -> np.ndarray:
-        rng = check_random_state(self.random_state)
-        return np.array(
-            [
-                self.classes_[int(rng.choice(np.flatnonzero(prob == prob.max())))]
-                for prob in self._fit_predict_proba(X, y)
-            ]
-        )
+        X_t = self._fit_fp_shared(X, y)
+        return self._rotf.fit_predict(X_t, y)
 
     def _fit_predict_proba(self, X, y) -> np.ndarray:
-        Xt = self._fit_fresh_prince(X, y, save_rotf_data=True)
-        return self._rotf._get_train_probs(Xt, y)
+        X_t = self._fit_fp_shared(X, y)
+        return self._rotf.fit_predict_proba(X_t, y)
 
-    def _fit_fresh_prince(self, X, y, save_rotf_data=False):
+    def _fit_fp_shared(self, X, y):
         self.n_cases_, self.n_channels_, self.n_timepoints_ = X.shape
 
         self._rotf = RotationForestClassifier(
             n_estimators=self.n_estimators,
-            save_transformed_data=save_rotf_data,
             n_jobs=self._n_jobs,
             random_state=self.random_state,
         )
@@ -184,10 +177,7 @@ def _fit_fresh_prince(self, X, y, save_rotf_data=False):
             disable_progressbar=self.verbose < 1,
         )
 
-        X_t = self._tsfresh.fit_transform(X, y)
-        self._rotf.fit(X_t, y)
-
-        return X_t
+        return self._tsfresh.fit_transform(X, y)
 
     @classmethod
     def get_test_params(cls, parameter_set="default"):

aeon/classification/shapelet_based/_stc.py

@@ -187,8 +187,8 @@ def _fit(self, X, y):
         Changes state by creating a fitted model that updates attributes
         ending in "_".
         """
-        self._fit_stc(X, y)
-
+        X_t = self._fit_stc_shared(X, y)
+        self._estimator.fit(X_t, y)
         return self
 
     def _predict(self, X) -> np.ndarray:
@@ -243,13 +243,15 @@ def _fit_predict(self, X, y) -> np.ndarray:
         )
 
     def _fit_predict_proba(self, X, y) -> np.ndarray:
-        Xt = self._fit_stc(X, y, save_rotf_data=True)
+        X_t = self._fit_stc_shared(X, y)
 
         if (isinstance(self.estimator, RotationForestClassifier)) or (
             self.estimator is None
         ):
-            return self._estimator._get_train_probs(Xt, y)
+            return self._estimator.fit_predict_proba(X_t, y)
         else:
+            self._estimator.fit(X_t, y)
+
             m = getattr(self._estimator, "predict_proba", None)
             if not callable(m):
                 raise ValueError("Estimator must have a predict_proba method.")
@@ -269,14 +271,14 @@ def _fit_predict_proba(self, X, y) -> np.ndarray:
 
             return cross_val_predict(
                 estimator,
-                X=Xt,
+                X=X_t,
                 y=y,
                 cv=cv_size,
                 method="predict_proba",
                 n_jobs=self._n_jobs,
             )
 
-    def _fit_stc(self, X, y, save_rotf_data=False):
+    def _fit_stc_shared(self, X, y):
         self.n_cases_, self.n_channels_, self.n_timepoints_ = X.shape
 
         if self.time_limit_in_minutes > 0:
@@ -304,9 +306,6 @@ def _fit_stc(self, X, y, save_rotf_data=False):
             self.random_state,
         )
 
-        if isinstance(self._estimator, RotationForestClassifier):
-            self._estimator.save_transformed_data = save_rotf_data
-
         m = getattr(self._estimator, "n_jobs", None)
         if m is not None:
             self._estimator.n_jobs = self._n_jobs
@@ -315,11 +314,7 @@ def _fit_stc(self, X, y, save_rotf_data=False):
         if m is not None and self.time_limit_in_minutes > 0:
             self._estimator.time_limit_in_minutes = self._classifier_limit_in_minutes
 
-        Xt = self._transformer.fit_transform(X, y)
-
-        self._estimator.fit(Xt, y)
-
-        return Xt
+        return self._transformer.fit_transform(X, y)
 
     @classmethod
     def get_test_params(cls, parameter_set="default"):