def distance(instance_a, instance_b, **params):
# find distance
instance_a = from_nested_to_2d_array(
instance_a, return_numpy=True) # todo use specific
# dimension rather than whole
# thing?
instance_b = from_nested_to_2d_array(
instance_b, return_numpy=True) # todo use specific
# dimension rather than whole thing?
instance_a = np.transpose(instance_a)
instance_b = np.transpose(instance_b)
return distance_measure(instance_a, instance_b, **params)
def test_row_transformer_transform_inverse_transform():
X, y = load_gunpoint(return_X_y=True)
t = RowTransformer(StandardScaler())
Xt = t.fit_transform(X)
Xit = t.inverse_transform(Xt)
assert Xit.shape == X.shape
assert isinstance(
Xit.iloc[0, 0], (pd.Series, np.ndarray)
) # check series-to-series transforms
np.testing.assert_array_almost_equal(
from_nested_to_2d_array(X).values,
from_nested_to_2d_array(Xit).values,
decimal=5,
)
def test_from_nested_to_2d_array(n_instances, n_columns, n_timepoints):
nested, _ = make_classification_problem(n_instances, n_columns,
n_timepoints)
array = from_nested_to_2d_array(nested)
assert array.shape == (n_instances, n_columns * n_timepoints)
assert array.index.equals(nested.index)
def transform(self, X, y=None):
"""Concatenate multivariate time series/panel data into long
univariate time series/panel
data by simply concatenating times series in time.
Parameters
----------
X : nested pandas DataFrame of shape [n_samples, n_features]
Nested dataframe with time-series in cells.
Returns
-------
Xt : pandas DataFrame
Transformed pandas DataFrame with same number of rows and single
column
"""
self.check_is_fitted()
X = check_X(X)
# We concatenate by tabularizing all columns and then detabularizing
# them into a single column
if isinstance(X, pd.DataFrame):
Xt = from_nested_to_2d_array(X)
else:
Xt = from_3d_numpy_to_2d_array(X)
return from_2d_array_to_nested(Xt)
def test_output_format_dim(len_series, n_instances, n_components):
np.random.seed(42)
X = from_2d_array_to_nested(
pd.DataFrame(data=np.random.randn(n_instances, len_series)))
trans = PCATransformer(n_components=n_components)
Xt = trans.fit_transform(X)
# Check number of rows and output type.
assert isinstance(Xt, pd.DataFrame)
assert Xt.shape[0] == X.shape[0]
# Check number of principal components in the output.
assert from_nested_to_2d_array(Xt).shape[1] == min(
n_components,
from_nested_to_2d_array(X).shape[1])
def _combine_data_frames(self, dataFrames, weighting_factor, col_names):
"""
Helper function for the shape_dtw class to combine two dataframes
together into a single dataframe.
Used when the shape_descriptor_function is set to "compound".
"""
first_desc = dataFrames[0]
second_desc = dataFrames[1]
first_desc_array = []
second_desc_array = []
# Convert the dataframes into arrays
for x in first_desc.columns:
first_desc_array.append(
from_nested_to_2d_array(first_desc[x], return_numpy=True)
)
for x in second_desc.columns:
second_desc_array.append(
from_nested_to_2d_array(second_desc[x], return_numpy=True)
)
# Concatenate the arrays together
res = []
for x in range(len(first_desc_array)):
dim1 = []
for y in range(len(first_desc_array[x])):
dim2 = []
dim2.extend(first_desc_array[x][y])
dim2.extend(second_desc_array[x][y] * weighting_factor)
dim1.append(dim2)
res.append(dim1)
res = np.asarray(res)
# Convert to pandas dataframe
df = pd.DataFrame()
for col in col_names:
colToAdd = []
for row in range(len(res[col])):
inst = res[col][row]
colToAdd.append(pd.Series(inst))
df[col] = colToAdd
return df
def _handle_tabularizer_args(*args, **kwargs):
# the Tabularizer transforms a nested pd.DataFrame/3d numpy array into a
# 2d numpy array, so the inverse transform goes from a 2d numpy array to a
# nested pd.DataFrame/3d array
# TODO refactor Tabularizer as series-as-features composition meta-estimator,
# rather than transformer or introduce special transformer type
X, y = args
if "return_numpy" in kwargs and kwargs["return_numpy"]:
return from_3d_numpy_to_2d_array(X), y
else:
return from_nested_to_2d_array(X), y
def row_first(X):
if isinstance(X, pd.Series):
X = pd.DataFrame(X)
Xt = pd.concat(
[
pd.Series(from_nested_to_2d_array(col).iloc[:, 0])
for _, col in X.items()
],
axis=1,
)
return Xt
def test_tsfresh_extractor(default_fc_parameters):
X, y = make_classification_problem()
X_train, X_test, y_train, y_test = train_test_split(X, y)
transformer = TSFreshFeatureExtractor(
default_fc_parameters=default_fc_parameters, disable_progressbar=True)
Xt = transformer.fit_transform(X_train, y_train)
actual = Xt.filter(like="__mean", axis=1).values.ravel()
expected = from_nested_to_2d_array(X_train).mean(axis=1).values
assert expected[0] == X_train.iloc[0, 0].mean()
np.testing.assert_allclose(actual, expected)
def _apply_rowwise(self, func, X, y=None):
"""Helper function to apply transform or inverse_transform function
on each row of data container"""
self.check_is_fitted()
X = check_X(X, coerce_to_pandas=True)
# 1st attempt: apply, relatively fast but not robust
# try and except, but sometimes breaks in other cases than excepted
# ValueError
# Works on single column, but on multiple columns only if columns
# have equal-length series.
# try:
# Xt = X.apply(self.transformer.fit_transform)
#
# # Otherwise call apply on each column separately.
# except ValueError as e:
# if str(e) == "arrays must all be same length":
# Xt = pd.concat([pd.Series(col.apply(
# self.transformer.fit_transform)) for _, col in X.items()],
# axis=1)
# else:
# raise
# 2nd attempt: apply but iterate over columns, still relatively fast
# but still not very robust
# but column is not 2d and thus breaks if transformer expects 2d input
try:
Xt = pd.concat(
[pd.Series(col.apply(func)) for _, col in X.items()], axis=1)
# 3rd attempt: explicit for-loops, most robust but very slow
except Exception:
cols_t = []
for c in range(X.shape[1]): # loop over columns
col = X.iloc[:, c]
rows_t = []
for row in col: # loop over rows in each column
row_2d = pd.DataFrame(row) # convert into 2d dataframe
row_t = func(row_2d).ravel() # apply transform
rows_t.append(row_t) # append transformed rows
cols_t.append(rows_t) # append transformed columns
# if series-to-series transform, flatten transformed series
Xt = _concat_nested_arrays(
cols_t) # concatenate transformed columns
# tabularise/unnest series-to-primitive transforms
xt = Xt.iloc[0, 0]
if isinstance(xt, (pd.Series, np.ndarray)) and len(xt) == 1:
Xt = from_nested_to_2d_array(Xt)
return Xt
def transform(self, X, y=None):
"""
Function to transform a data frame of time series data.
Parameters
----------
X : a pandas dataframe of shape = [n_samples, num_dims]
The training input samples.
Returns
-------
dims: a pandas data frame of shape = [n_samples, num_dims]
"""
# Check the data
self.check_is_fitted()
X = check_X(X, enforce_univariate=False, coerce_to_pandas=True)
# Get information about the dataframe
num_insts = X.shape[0]
col_names = X.columns
num_atts = len(X.iloc[0, 0])
# Check the parameters are appropriate
self._check_parameters(num_atts)
df = pd.DataFrame()
for x in col_names:
# Convert one of the columns in the dataframe to a numpy array
arr = from_nested_to_2d_array(pd.DataFrame(X[x]),
return_numpy=True)
# Get the HOG1Ds of each time series
transformedData = []
for y in range(num_insts):
inst = self._calculate_hog1ds(arr[y])
transformedData.append(inst)
# Convert to numpy array
transformedData = np.asarray(transformedData)
# Add it to the dataframe
colToAdd = []
for i in range(len(transformedData)):
inst = transformedData[i]
colToAdd.append(pd.Series(inst))
df[x] = colToAdd
return df
def test_pca_results(n_components):
np.random.seed(42)
# sklearn
X = pd.DataFrame(data=np.random.randn(10, 5))
pca = PCA(n_components=n_components)
Xt1 = pca.fit_transform(X)
# sktime
Xs = from_2d_array_to_nested(X)
pca_transform = PCATransformer(n_components=n_components)
Xt2 = pca_transform.fit_transform(Xs)
assert np.allclose(np.asarray(Xt1), np.asarray(from_nested_to_2d_array(Xt2)))
def test_padding_transformer():
# load data
name = "JapaneseVowels"
X_train, y_train = _load_dataset(name, split="train", return_X_y=True)
X_test, y_test = _load_dataset(name, split="test", return_X_y=True)
# print(X_train)
padding_transformer = PaddingTransformer()
Xt = padding_transformer.fit_transform(X_train)
# when we tabulrize the data it has 12 dimensions
# and we've padded them to there normal length of 29
data = from_nested_to_2d_array(Xt)
assert len(data.columns) == 29 * 12
def test_truncation_paramterised_transformer():
# load data
name = "JapaneseVowels"
X_train, y_train = _load_dataset(name, split="train", return_X_y=True)
X_test, y_test = _load_dataset(name, split="test", return_X_y=True)
# print(X_train)
truncated_transformer = TruncationTransformer(2, 10)
Xt = truncated_transformer.fit_transform(X_train)
# when we tabulrize the data it has 12 dimensions
# and we've truncated them all to (10-2) long.
data = from_nested_to_2d_array(Xt)
assert len(data.columns) == 8 * 12
def test_padding_fill_value_transformer():
# load data
name = "JapaneseVowels"
X_train, y_train = _load_dataset(name, split="train", return_X_y=True)
X_test, y_test = _load_dataset(name, split="test", return_X_y=True)
# print(X_train)
padding_transformer = PaddingTransformer(pad_length=40, fill_value=1)
Xt = padding_transformer.fit_transform(X_train)
# when we tabulrize the data it has 12 dimensions
# and we've truncated them all to (10-2) long.
data = from_nested_to_2d_array(Xt)
assert len(data.columns) == 40 * 12
def transform(self, X, y=None):
"""
Parameters
----------
X : a pandas dataframe of shape = [n_samples, num_dims]
The training input samples.
Returns
-------
df: a pandas data frame of shape = [num_intervals, num_dims]
"""
# Check the data
self.check_is_fitted()
X = check_X(X, coerce_to_pandas=True)
# Get information about the dataframe
n_timepoints = len(X.iloc[0, 0])
num_instances = X.shape[0]
col_names = X.columns
self._check_parameters(n_timepoints)
df = pd.DataFrame()
for x in col_names:
# Convert one of the columns in the dataframe to numpy array
arr = from_nested_to_2d_array(pd.DataFrame(X[x]), return_numpy=True)
# Calculate gradients
transformedData = []
for y in range(num_instances):
res = self._get_gradients_of_lines(arr[y])
transformedData.append(res)
# Convert to Numpy array
transformedData = np.asarray(transformedData)
# Add it to the dataframe
colToAdd = []
for i in range(len(transformedData)):
inst = transformedData[i]
colToAdd.append(pd.Series(inst))
df[x] = colToAdd
return df
def _perform_paa_along_dim(self, X):
X = from_nested_to_2d_array(X, return_numpy=True)
num_atts = X.shape[1]
num_insts = X.shape[0]
dims = pd.DataFrame()
data = []
for i in range(num_insts):
series = X[i, :]
frames = []
current_frame = 0
current_frame_size = 0
frame_length = num_atts / self.num_intervals
frame_sum = 0
for n in range(num_atts):
remaining = frame_length - current_frame_size
if remaining > 1:
frame_sum += series[n]
current_frame_size += 1
else:
frame_sum += remaining * series[n]
current_frame_size += remaining
if current_frame_size == frame_length:
frames.append(frame_sum / frame_length)
current_frame += 1
frame_sum = (1 - remaining) * series[n]
current_frame_size = 1 - remaining
# if the last frame was lost due to double imprecision
if current_frame == self.num_intervals - 1:
frames.append(frame_sum / frame_length)
data.append(pd.Series(frames))
dims[0] = data
return dims
def transform(self, X, y=None):
"""
Parameters
----------
X : a pandas dataframe of shape = [n_samples, num_dims]
The training input samples.
Returns
-------
dims: a pandas data frame of shape
= [n_samples, num_dims]
"""
# Check the data
self.check_is_fitted()
X = check_X(X, enforce_univariate=False, coerce_to_pandas=True)
self._check_parameters()
# Get information about the dataframe
col_names = X.columns
df = pd.DataFrame()
for x in col_names:
# Convert one of the columns in the dataframe to numpy array
arr = from_nested_to_2d_array(pd.DataFrame(X[x]),
return_numpy=True)
transformedData = self._extract_wavelet_coefficients(arr)
# Convert to a numpy array
transformedData = np.asarray(transformedData)
# Add it to the dataframe
colToAdd = []
for i in range(len(transformedData)):
inst = transformedData[i]
colToAdd.append(pd.Series(inst))
df[x] = colToAdd
return df
def transform(self, X, y=None):
"""Transform nested pandas dataframe into tabular dataframe.
Parameters
----------
X : pandas DataFrame
Nested dataframe with pandas series or numpy arrays in cells.
y : array-like, optional (default=None)
Returns
-------
Xt : pandas DataFrame
Transformed dataframe with only primitives in cells.
"""
self.check_is_fitted()
X = check_X(X)
if isinstance(X, pd.DataFrame):
return from_nested_to_2d_array(X)
else:
return from_3d_numpy_to_2d_array(X)
def test_dft_mft():
# load training data
X, Y = load_gunpoint(split="train", return_X_y=True)
X_tab = from_nested_to_2d_array(X, return_numpy=True)
word_length = 6
alphabet_size = 4
# print("Single DFT transformation")
window_size = np.shape(X_tab)[1]
p = SFA(
word_length=word_length,
alphabet_size=alphabet_size,
window_size=window_size,
binning_method="equi-depth",
).fit(X, Y)
dft = p._discrete_fourier_transform(X_tab[0])
mft = p._mft(X_tab[0])
assert (mft - dft < 0.0001).all()
# print("Windowed DFT transformation")
for norm in [True, False]:
for window_size in [140]:
p = SFA(
word_length=word_length,
norm=norm,
alphabet_size=alphabet_size,
window_size=window_size,
binning_method="equi-depth",
).fit(X, Y)
mft = p._mft(X_tab[0])
for i in range(len(X_tab[0]) - window_size + 1):
dft_transformed = p._discrete_fourier_transform(
X_tab[0, i : window_size + i]
)
assert (mft[i] - dft_transformed < 0.001).all()
assert len(mft) == len(X_tab[0]) - window_size + 1
assert len(mft[0]) == word_length
def _transform_single_feature(self, X, feature):
"""transforms data into the catch22 features
Parameters
----------
X : pandas DataFrame, input time series
feature : int, catch22 feature id or String, catch22 feature
name.
Returns
-------
Numpy array containing a catch22 feature for each input series
"""
if isinstance(feature, int):
if feature > 21 or feature < 0:
raise ValueError("Invalid catch22 feature ID")
elif isinstance(feature, str):
if feature in feature_names:
feature = feature_names.index(feature)
else:
raise ValueError("Invalid catch22 feature name")
else:
raise ValueError("Feature name or ID required")
if isinstance(X, pd.DataFrame):
X = from_nested_to_2d_array(X, return_numpy=True)
n_instances = X.shape[0]
X = np.reshape(X, (n_instances, -1))
c22_list = []
for i in range(n_instances):
series = X[i, :].tolist()
c22_val = features[feature](series)
c22_list.append(c22_val)
return np.array(c22_list)
