Skip to content

[ENH] add unequal-length time series support for tsfresh-based methods #3187

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
jsquaredosquared wants to merge 8 commits into
base: main
Choose a base branch
from
+68 −26
Open

[ENH] add unequal-length time series support for tsfresh-based methods #3187

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
07ecfc9
feat: add unequal-length time series support for tsfresh-based methods
jsquaredosquared Dec 20, 2025
5463ccd
docs: modify docstrings to be consistent
jsquaredosquared Dec 21, 2025
ba70c76
update x_inner_type from numpy3D to np-list
jsquaredosquared Dec 21, 2025
44d1a24
fix: add missing X_inner_type tag
jsquaredosquared Dec 20, 2025
b86ac6f
Merge branch 'main' into tsfresh-unequal
jsquaredosquared Dec 22, 2025
593cfa3
Merge branch 'main' into tsfresh-unequal
jsquaredosquared Dec 29, 2025
4401568
test: include unequal length time series in tsfresh extractor test
jsquaredosquared Feb 2, 2026
e4829fd
Merge branch 'main' into tsfresh-unequal
jsquaredosquared Feb 2, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions aeon/classification/feature_based/_tsfresh.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,8 +83,10 @@ class TSFreshClassifier(BaseClassifier):
"""

_tags = {
"X_inner_type": ["np-list", "numpy3D"],
"capability:multivariate": True,
"capability:multithreading": True,
"capability:unequal_length": True,
"algorithm_type": "feature",
"python_dependencies": "tsfresh",
}
Expand Down Expand Up @@ -122,6 +124,8 @@ def _fit(self, X, y):
Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The training data.
y : array-like, shape = [n_cases]
The class labels.
Expand Down Expand Up @@ -193,6 +197,8 @@ def _predict(self, X) -> np.ndarray:
Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The data to make predictions for.

Returns
Expand All @@ -211,6 +217,8 @@ def _predict_proba(self, X) -> np.ndarray:
Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The data to make predict probabilities for.

Returns
Expand Down
8 changes: 8 additions & 0 deletions aeon/clustering/feature_based/_tsfresh.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,8 +71,10 @@ class TSFreshClusterer(BaseClusterer):
"""

_tags = {
"X_inner_type": ["np-list", "numpy3D"],
"capability:multivariate": True,
"capability:multithreading": True,
"capability:unequal_length": True,
"algorithm_type": "feature",
"python_dependencies": "tsfresh",
}
Expand Down Expand Up @@ -107,6 +109,8 @@ def _fit(self, X: np.ndarray, y: np.ndarray | None = None):
Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The training data.
y : array-like, shape = [n_cases]
Ignored. The class labels.
Expand Down Expand Up @@ -172,6 +176,8 @@ def _predict(self, X: np.ndarray) -> np.ndarray:
Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The data to make predictions for.

Returns
Expand All @@ -187,6 +193,8 @@ def _predict_proba(self, X: np.ndarray) -> np.ndarray:
Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The data to make predictions for.

Returns
Expand Down
10 changes: 8 additions & 2 deletions aeon/regression/feature_based/_tsfresh.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,8 +64,10 @@ class TSFreshRegressor(BaseRegressor):
"""

_tags = {
"X_inner_type": ["np-list", "numpy3D"],
"capability:multivariate": True,
"capability:multithreading": True,
"capability:unequal_length": True,
"algorithm_type": "feature",
"python_dependencies": "tsfresh",
}
Expand Down Expand Up @@ -101,7 +103,9 @@ def _fit(self, X, y):

Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
X : 3D np.ndarray of shape [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The training data.
y : array-like, shape = [n_cases]
The target labels.
Expand Down Expand Up @@ -172,7 +176,9 @@ def _predict(self, X) -> np.ndarray:

Parameters
----------
X : 3D np.ndarray of shape = [n_cases, n_channels, n_timepoints]
X : 3D np.ndarray of shape [n_cases, n_channels, n_timepoints]
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
The data to make predictions for.

Returns
Expand Down
50 changes: 30 additions & 20 deletions aeon/transformations/collection/feature_based/_tsfresh.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,22 +10,27 @@
from aeon.utils.validation import check_n_jobs


def _from_3d_numpy_to_long(arr):
# Converting the numpy array to a long format DataFrame
n_cases, n_channels, n_timepoints = arr.shape

# Creating a DataFrame from the numpy array with multi-level index
df = pd.DataFrame(arr.reshape(n_cases * n_channels, n_timepoints))
df["case_index"] = np.repeat(np.arange(n_cases), n_channels)
df["dimension"] = np.tile(np.arange(n_channels), n_cases)
df = df.melt(
id_vars=["case_index", "dimension"], var_name="time_index", value_name="value"
def _from_collection_to_long(collection):
n_cases = len(collection)
n_channels = collection[0].shape[0]
n_timepoints = np.array([arr.shape[1] for arr in collection])

index = np.repeat(np.arange(n_cases), n_channels * n_timepoints)
timepoints = [np.arange(timepoints_i) for timepoints_i in n_timepoints]
time_index = np.concatenate([np.tile(arr, n_channels) for arr in timepoints])
column = np.concatenate(
[
np.repeat(np.arange(n_channels), timepoints_i)
for timepoints_i in n_timepoints
]
)
value = np.concatenate([arr.flatten() for arr in collection])

# Adjusting the column order and renaming columns
df = df[["case_index", "time_index", "dimension", "value"]]
df = df.rename(columns={"case_index": "index", "dimension": "column"})
df = pd.DataFrame(
{"index": index, "time_index": time_index, "column": column, "value": value}
)
df["column"] = "dim_" + df["column"].astype(str)

return df


Expand All @@ -34,8 +39,10 @@ class _TSFresh(BaseCollectionTransformer):

_tags = {
"output_data_type": "Tabular",
"X_inner_type": ["np-list", "numpy3D"],
"capability:multivariate": True,
"capability:multithreading": True,
"capability:unequal_length": True,
"fit_is_empty": True,
"python_dependencies": "tsfresh",
}
Expand Down Expand Up @@ -278,17 +285,18 @@ def _transform(self, X, y=None):

Parameters
----------
X : 3D numpy array of shape (n_cases, n_channels, n_features)
X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
input time series collection.
y : ignored argument for interface compatibility

Returns
-------
X : 3D numpy array of shape (n_cases, n_channels, n_features)
input time series collection.
transformed version of X
Xt : np.ndarray of shape (n_cases, n_features)
Transformed data.
"""
Xt = _from_3d_numpy_to_long(X)
Xt = _from_collection_to_long(X)

# lazy imports to avoid hard dependency
from tsfresh import extract_features
Expand Down Expand Up @@ -345,7 +353,7 @@ def _get_names(self):
from tsfresh import extract_features

X = np.random.random((2, 1, 30))
Xt = _from_3d_numpy_to_long(X)
Xt = _from_collection_to_long(X)
Xt = extract_features(
Xt,
column_id="index",
Expand Down Expand Up @@ -475,7 +483,7 @@ class TSFreshRelevant(_TSFresh):

_tags = {
"requires_y": True,
"X_inner_type": "numpy3D",
"X_inner_type": ["np-list", "numpy3D"],
"fit_is_empty": False,
}

Expand Down Expand Up @@ -571,6 +579,8 @@ def _fit_transform(self, X, y=None):
Parameters
----------
X : 3D np.ndarray of shape (n_cases, n_channels, n_timepoints)
or list of np.ndarray of shape [n_cases], where each array is a
2D np.ndarray of shape = [n_channels, n_timepoints_i]
collection of time series to transform
y : Series, default=None
Additional data, e.g., labels for transformation
Expand Down