[ENH] Make transformer files private (#2072)

* remove Theta

* take some private

* import

* revert Interpolator

* softdep

* hog1D

* API

* Revert "API"

This reverts commit 428900b.

* API

* API

* API

* take collection transformers private

* docstring

* tidy

* remove unused function

* remove unused function

* examples

* docstrings

* revert things

aeon/transformations/collection/__init__.py

@@ -26,27 +26,27 @@
     "Normalise",
 ]
 
+from aeon.transformations.collection._acf import AutocorrelationFunctionTransformer
+from aeon.transformations.collection._ar_coefficient import ARCoefficientTransformer
 from aeon.transformations.collection._downsample import DownsampleTransformer
+from aeon.transformations.collection._dwt import DWTTransformer
+from aeon.transformations.collection._hog1d import HOG1DTransformer
+from aeon.transformations.collection._matrix_profile import MatrixProfile
 from aeon.transformations.collection._normalise import Normalise
 from aeon.transformations.collection._pad import Padder
+from aeon.transformations.collection._periodogram import PeriodogramTransformer
+from aeon.transformations.collection._reduce import Tabularizer
 from aeon.transformations.collection._resize import Resizer
+from aeon.transformations.collection._scaler import TimeSeriesScaler
+from aeon.transformations.collection._segment import (
+    IntervalSegmenter,
+    RandomIntervalSegmenter,
+    SlidingWindowSegmenter,
+)
+from aeon.transformations.collection._slope import SlopeTransformer
 from aeon.transformations.collection._truncate import Truncator
-from aeon.transformations.collection.acf import AutocorrelationFunctionTransformer
-from aeon.transformations.collection.ar_coefficient import ARCoefficientTransformer
 from aeon.transformations.collection.base import BaseCollectionTransformer
 from aeon.transformations.collection.channel_selection import (
     ElbowClassPairwise,
     ElbowClassSum,
 )
-from aeon.transformations.collection.dwt import DWTTransformer
-from aeon.transformations.collection.hog1d import HOG1DTransformer
-from aeon.transformations.collection.matrix_profile import MatrixProfile
-from aeon.transformations.collection.periodogram import PeriodogramTransformer
-from aeon.transformations.collection.reduce import Tabularizer
-from aeon.transformations.collection.scaler import TimeSeriesScaler
-from aeon.transformations.collection.segment import (
-    IntervalSegmenter,
-    RandomIntervalSegmenter,
-    SlidingWindowSegmenter,
-)
-from aeon.transformations.collection.slope import SlopeTransformer

aeon/transformations/collection/ar_coefficient.py → aeon/transformations/collection/_ar_coefficient.py

@@ -68,7 +68,7 @@ def _transform(self, X, y=None):
         order = self.order(X) if callable(self.order) else self.order
         if order > n_timepoints - self.min_values:
             order = n_timepoints - self.min_values
-        if order < 0:
+        if order <= 0:
             order = 1
 
         if order > n_timepoints - 1:

aeon/transformations/collection/dwt.py → aeon/transformations/collection/_dwt.py

@@ -4,7 +4,7 @@
 
 import numpy as np
 
-from aeon.transformations.collection import BaseCollectionTransformer
+from aeon.transformations.collection.base import BaseCollectionTransformer
 from aeon.utils.numba.wavelets import haar_transform
 
 

aeon/transformations/collection/hog1d.py → aeon/transformations/collection/_hog1d.py

@@ -5,7 +5,7 @@
 
 import numpy as np
 
-from aeon.transformations.collection import BaseCollectionTransformer
+from aeon.transformations.collection.base import BaseCollectionTransformer
 from aeon.utils import split_series
 
 
@@ -48,7 +48,7 @@ def _transform(self, X, y=None):
 
         Parameters
         ----------
-        X : 3D np.ndarray of shape = [n_cases, 1, n_timepoints]
+        X : 3D np.ndarray of shape (n_cases, 1, n_timepoints)
             collection of time series to transform
         y : ignored argument for interface compatibility
 
@@ -60,8 +60,6 @@ def _transform(self, X, y=None):
         """
         # Get information about the dataframe
         n_cases, n_channels, n_timepoints = X.shape
-        if n_channels > 1:
-            raise ValueError("HOG1D does not support multivariate time series.")
         # Check the parameters are appropriate
         self._check_parameters(n_timepoints)
         transX = []

aeon/transformations/collection/matrix_profile.py → aeon/transformations/collection/_matrix_profile.py

@@ -4,7 +4,7 @@
 
 import numpy as np
 
-from aeon.transformations.collection import BaseCollectionTransformer
+from aeon.transformations.collection.base import BaseCollectionTransformer
 
 
 def _sliding_dot_products(q, t, q_len, t_len):
@@ -217,8 +217,7 @@ def _transform(self, X, y=None):
 
         Parameters
         ----------
-        X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
-            panel of time series to transform
+        X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
         y : ignored argument for interface compatibility
 
         Returns

aeon/transformations/collection/periodogram.py → aeon/transformations/collection/_periodogram.py

@@ -7,14 +7,15 @@
 
 import numpy as np
 
-from aeon.transformations.collection import BaseCollectionTransformer
+from aeon.transformations.collection.base import BaseCollectionTransformer
 from aeon.utils.validation import check_n_jobs
 
 
 class PeriodogramTransformer(BaseCollectionTransformer):
     """Periodogram transformer.
 
-    This transformer converts a time series into its periodogram representation.
+    This transformer converts a collection of time series into its periodogram
+    representation.
 
     Parameters
     ----------
@@ -82,17 +83,15 @@ def _transform(self, X, y=None):
             kwargs = {"mode": self.pad_with}
             if self.pad_with == "constant":
                 kwargs["constant_values"] = self.constant_value
-
+            len = int(math.pow(2, math.ceil(math.log(X.shape[2], 2))) - X.shape[2])
             X = np.pad(
                 X,
                 (
                     (0, 0),
                     (0, 0),
                     (
                         0,
-                        int(
-                            math.pow(2, math.ceil(math.log(X.shape[2], 2))) - X.shape[2]
-                        ),
+                        len,
                     ),
                 ),
                 **kwargs,

aeon/transformations/collection/reduce.py → aeon/transformations/collection/_reduce.py

@@ -3,7 +3,7 @@
 __maintainer__ = []
 __all__ = ["Tabularizer"]
 
-from aeon.transformations.collection import BaseCollectionTransformer
+from aeon.transformations.collection.base import BaseCollectionTransformer
 
 
 class Tabularizer(BaseCollectionTransformer):
@@ -23,12 +23,12 @@ class Tabularizer(BaseCollectionTransformer):
     }
 
     def _transform(self, X, y=None):
-        """Transform nested pandas dataframe into tabular dataframe.
+        """Transform collection into tabular dataframe.
 
         Parameters
         ----------
-        X : pandas DataFrame or 3D np.ndarray
-            panel of time series to transform
+        X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
+            Collection of time series to transform.
         y : ignored argument for interface compatibility
 
         Returns

aeon/transformations/collection/segment.py → aeon/transformations/collection/_segment.py

@@ -6,7 +6,7 @@
 import pandas as pd
 from sklearn.utils import check_random_state
 
-from aeon.transformations.collection import BaseCollectionTransformer
+from aeon.transformations.collection.base import BaseCollectionTransformer
 from aeon.utils.datetime import get_time_index
 from aeon.utils.validation import check_window_length
 
@@ -407,7 +407,7 @@ class SlidingWindowSegmenter(BaseCollectionTransformer):
     Examples
     --------
     >>> from aeon.datasets import load_unit_test
-    >>> from aeon.transformations.collection.segment import SlidingWindowSegmenter
+    >>> from aeon.transformations.collection import SlidingWindowSegmenter
     >>> data = np.array([[[1, 2, 3, 4, 5, 6, 7, 8]], [[5, 5, 5, 5, 5, 5, 5, 5]]])
     >>> seggy = SlidingWindowSegmenter(window_length=4)
     >>> data2 = seggy.fit_transform(data)

aeon/transformations/collection/slope.py → aeon/transformations/collection/_slope.py

@@ -7,7 +7,7 @@
 
 import numpy as np
 
-from aeon.transformations.collection import BaseCollectionTransformer
+from aeon.transformations.collection.base import BaseCollectionTransformer
 from aeon.utils import split_series
 
 
@@ -29,7 +29,7 @@ class SlopeTransformer(BaseCollectionTransformer):
     Examples
     --------
     >>> import numpy as np
-    >>> from aeon.transformations.collection.slope import SlopeTransformer
+    >>> from aeon.transformations.collection import SlopeTransformer
     >>> X = np.array([[[4, 6, 10, 12, 8, 6, 5, 5]]])
     >>> s = SlopeTransformer(n_intervals=2)
     >>> res = s.fit_transform(X)
@@ -57,8 +57,7 @@ def _transform(self, X, y=None):
 
         Returns
         -------
-        3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
-        collection of time series to transform
+        3D np.ndarray of shape = [n_cases, n_channels, n_intervals]
         """
         # Get information about the dataframe
         n_cases, n_channels, n_timepoints = X.shape

aeon/transformations/collection/_truncate.py

@@ -1,4 +1,4 @@
-"""Truncation transformer - truncate unequal length panels to lower/upper bounds."""
+"""Truncation transformer - truncate unequal length collections."""
 
 __all__ = ["Truncator"]
 __maintainer__ = []
@@ -11,9 +11,9 @@
 class Truncator(BaseCollectionTransformer):
     """Truncate unequal length time series to a lower bounds.
 
-    Truncates all series in panel between lower/upper range bounds. This transformer
-    assumes that all series have the same number of channels (dimensions) and
-    that all channels in a single series are the same length.
+    Truncates all series in collection between lower/upper range bounds. This
+    transformer assumes that all series have the same number of channels (dimensions)
+    and that all channels in a single series are the same length.
 
     Parameters
     ----------

aeon/transformations/collection/base.py

@@ -377,20 +377,21 @@ def _update(self, X, y=None):
         # standard behaviour: no update takes place, new data is ignored
         return self
 
-    def _check_y(self, y, n_cases):
-        if y is None:
-            return None
-        # Check y valid input for collection transformations
-        if not isinstance(y, (pd.Series, np.ndarray)):
-            raise TypeError(
-                f"y must be a np.array or a pd.Series, but found type: {type(y)}"
-            )
-        if isinstance(y, np.ndarray) and y.ndim > 1:
-            raise TypeError(f"y must be 1-dimensional, found {y.ndim} dimensions")
-        # Check matching number of labels
-        n_labels = y.shape[0]
-        if n_cases != n_labels:
-            raise ValueError(
-                f"Mismatch in number of cases. Number in X = {n_cases} nos in y = "
-                f"{n_labels}"
-            )
+
+def _check_y(self, y, n_cases):
+    if y is None:
+        return None
+    # Check y valid input for collection transformations
+    if not isinstance(y, (pd.Series, np.ndarray)):
+        raise TypeError(
+            f"y must be a np.array or a pd.Series, but found type: {type(y)}"
+        )
+    if isinstance(y, np.ndarray) and y.ndim > 1:
+        raise TypeError(f"y must be 1-dimensional, found {y.ndim} dimensions")
+    # Check matching number of labels
+    n_labels = y.shape[0]
+    if n_cases != n_labels:
+        raise ValueError(
+            f"Mismatch in number of cases. Number in X = {n_cases} nos in y = "
+            f"{n_labels}"
+        )

aeon/transformations/collection/convolution_based/_minirocket.py

@@ -96,8 +96,8 @@ def _fit(self, X, y=None):
 
         Parameters
         ----------
-        X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
-            panel of time series to transform
+        X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
+            Collection of time series to transform
         y : ignored argument for interface compatibility
 
         Returns
@@ -128,8 +128,8 @@ def _transform(self, X, y=None):
 
         Parameters
         ----------
-        X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
-            panel of time series to transform
+        X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
+            Collection of time series to transform
         y : ignored argument for interface compatibility
 
         Returns

aeon/transformations/collection/convolution_based/_multirocket.py

@@ -105,8 +105,8 @@ def _fit(self, X, y=None):
 
         Parameters
         ----------
-        X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
-            panel of time series to transform
+        X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
+            Collection of time series to transform
         y : ignored argument for interface compatibility
 
         Returns
@@ -147,8 +147,8 @@ def _transform(self, X, y=None):
 
         Parameters
         ----------
-        X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
-            panel of time series to transform
+        X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
+            Collection of time series to transform.
         y : ignored argument for interface compatibility
 
         Returns

aeon/transformations/collection/feature_based/_tsfresh.py

@@ -550,19 +550,12 @@ def _fit_transform(self, X, y=None):
 
         Writes to self:
         _is_fitted : flag is set to True.
-        _X : X, coerced copy of X, if remember_data tag is True
-            possibly coerced to inner type or update_data compatible type
-            by reference, when possible
-        model attributes (ending in "_") : dependent on estimator
 
         Parameters
         ----------
-        X : Series or Panel, any supported type
-            Data to be transformed, of python type as follows:
-                Series: pd.Series, pd.DataFrame, or np.ndarray (1D or 2D)
-                Panel: pd.DataFrame with 2-level MultiIndex, list of pd.DataFrame,
-                    nested pd.DataFrame, or pd.DataFrame in long/wide format
-        y : Series or Panel, default=None
+        X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
+            collection of time series to transform
+        y : Series, default=None
             Additional data, e.g., labels for transformation
 
         Returns

aeon/transformations/collection/signature_based/__init__.py

@@ -1,4 +1,4 @@
-"""Signature based panel transformations."""
+"""Signature based Collection transformations."""
 
 __all__ = [
     "SignatureTransformer",

aeon/transformations/collection/tests/test_acf.py

@@ -0,0 +1,14 @@
+"""Test for incorrect input for ARCoefficientTransformer transformer."""
+
+import numpy as np
+import pytest
+
+from aeon.transformations.collection import AutocorrelationFunctionTransformer
+
+
+def test_acf():
+    """Test ACF Transformer exceptions."""
+    X = np.random.random((2, 1, 10))
+    acf = AutocorrelationFunctionTransformer(n_lags=100)
+    with pytest.raises(ValueError, match=r"must be smaller than n_timepoints - 1"):
+        acf.fit_transform(X)

aeon/transformations/collection/tests/test_ar_coefficient.py

@@ -0,0 +1,28 @@
+"""Test for incorrect input for ARCoefficientTransformer transformer."""
+
+import numpy as np
+import pytest
+
+from aeon.transformations.collection import ARCoefficientTransformer
+from aeon.utils.validation._dependencies import _check_soft_dependencies
+
+
+@pytest.mark.skipif(
+    not _check_soft_dependencies("statsmodels", severity="none"),
+    reason="skip test if required soft dependency statsmodels not available",
+)
+def test_ar_coefficient():
+    """Test AR Coefficient Transformer exceptions."""
+    X = np.random.random((2, 1, 10))
+    ar = ARCoefficientTransformer(order=0)
+    Xt = ar.fit_transform(X)
+    assert Xt.shape[2] == 1
+    ar = ARCoefficientTransformer(order=-10)
+    Xt = ar.fit_transform(X)
+    assert Xt.shape[2] == 1
+    ar = ARCoefficientTransformer(order=100)
+    with pytest.raises(ValueError, match=r"must be smaller than n_timepoints - 1"):
+        ar.fit_transform(X)
+    ar = ARCoefficientTransformer(order=6, min_values=5)
+    Xt = ar.fit_transform(X)
+    assert Xt.shape[2] == 5