def test_shuffle_make_cylinder_bell_funnel(params):
"""Test that shuffling works as expected."""
_, y = make_cylinder_bell_funnel(**params)
arr_desired_no_shuffle = np.repeat(np.arange(3), 10)
if 'shuffle' in params.keys() and (not params['shuffle']):
np.testing.assert_array_equal(y, arr_desired_no_shuffle)
else:
assert not np.array_equal(y, arr_desired_no_shuffle)
def test_return_params_make_cylinder_bell_funnel(params):
"""Test the return objects."""
res = make_cylinder_bell_funnel(**params)
assert isinstance(res[0], np.ndarray)
assert isinstance(res[1], np.ndarray)
if 'return_params' in params.keys() and params['return_params']:
parameters = res[2]
assert isinstance(parameters, dict)
assert isinstance(parameters['a'], (int, np.integer))
assert isinstance(parameters['b'], (int, np.integer))
assert isinstance(parameters['eta'], (float, np.floating))
assert isinstance(parameters['epsilon'], np.ndarray)
if 'n_samples' in params:
assert parameters['epsilon'].shape == (params['n_samples'], 128)
else:
assert parameters['epsilon'].shape == (30, 128)
It is implemented as :class:`pyts.decomposition.SingularSpectrumAnalysis`.
"""
# Author: Lucas Plagwitz <*****@*****.**>
# License: BSD-3-Clause
import numpy as np
import matplotlib.pyplot as plt
from pyts.decomposition import SingularSpectrumAnalysis
from pyts.datasets import make_cylinder_bell_funnel
# Parameters
n_samples, n_timestamps = 3, 128
X_cbf, y = make_cylinder_bell_funnel(n_samples=10,
random_state=42,
shuffle=False)
X_period = 3 * np.sin(np.arange(n_timestamps))
X = X_cbf[:, :n_timestamps] + X_period
# We decompose the time series into three subseries
window_size = 20
# Singular Spectrum Analysis
ssa = SingularSpectrumAnalysis(window_size=window_size, groups="auto")
X_ssa = ssa.fit_transform(X)
# Show the results for different frequency-parameters
plt.figure(figsize=(16, 12))
ax1 = plt.subplot(221)
def rolling_window(a, window, step_size):
shape = (np.int((a.shape[-1] - window) / step_size) + 1, window)
rolled = np.empty(shape)
for i in np.arange(shape[0]):
rolled[i, :] = a[i * step_size:i * step_size + window]
return rolled
# %% Cylinder_bell_funnel data
number_of_samples = 12
data, labels = make_cylinder_bell_funnel(n_samples=number_of_samples,
random_state=0,
shuffle=True,
weights=(1 / 2, 1 / 2))
fs = 1
length_of_window = 128
timeseries = np.hstack(data)
np.savetxt('cylinderbellfunneltimeseries.csv',
np.concatenate((np.reshape(np.arange(timeseries.size),
(timeseries.size, 1)),
np.reshape(timeseries, (timeseries.size, 1))),
axis=1),
delimiter=',')
windowed_data = rolling_window(timeseries, length_of_window, 1)
def test_parameter_check_make_cylinder_bell_funnel(params, error, err_msg):
"""Test parameter validation."""
with pytest.raises(error, match=re.escape(err_msg)):
make_cylinder_bell_funnel(**params)
def test_class_balance_make_cylinder_bell_funnel(params,
class_balance_desired):
"""Test that the class balance is the expected one."""
X, y = make_cylinder_bell_funnel(**params)
class_balance_actual = np.bincount(y)
np.testing.assert_array_equal(class_balance_actual, class_balance_desired)
return rolled
# %% Cylinder_bell_funnel data --> CHOOSE WHETHER YOU WANT TWO OR THREE CLUSTERS
number_of_samples = 12
proportion = (1 / 2, 1 / 2) #For only cylinder and bell shapes
amount_of_clusters = 2
#proportion = (1/3,1/3) #For all three shapes
#amount_of_clusters = 3
data, labels = make_cylinder_bell_funnel(n_samples=number_of_samples,
shuffle=True,
weights=proportion)
fs = 1
length_of_window = 128
timeseries = np.hstack(data)
np.savetxt('cylinderbellfunneltimeseries.csv',
np.concatenate((np.reshape(np.arange(timeseries.size),
(timeseries.size, 1)),
np.reshape(timeseries, (timeseries.size, 1))),
axis=1),
delimiter=',')
windowed_data = rolling_window(timeseries, length_of_window, 1)
"""
=====================================
Making a Cylinder-Bell-Funnel dataset
=====================================
This example shows how to generate a Cylinder-Bell-Funnel dataset. This
simulated dataset was introduced by N. Saito in his Ph.D. thesis entitled
"Local feature extraction and its application". It is one of the most
well-known datasets in time series classification.
It is implemented as :func:`pyts.datasets.make_cylinder_bell_funnel`.
"""
# Author: Johann Faouzi <*****@*****.**>
# License: BSD-3-Clause
import matplotlib.pyplot as plt
from pyts.datasets import make_cylinder_bell_funnel
X, y = make_cylinder_bell_funnel(n_samples=12, random_state=42)
plt.figure(figsize=(12, 9))
for i, classe in enumerate(['cylinder', 'bell', 'funnel']):
plt.subplot(3, 1, i + 1)
for x in X[y == i]:
plt.plot(x, color='C0', linewidth=0.9)
plt.title('Class: {}'.format(classe), fontsize=16)
plt.tight_layout()
plt.subplots_adjust(hspace=0.4)
plt.show()
color=color,
**kwargs)
return self
if __name__ == '__main__':
from pyts.datasets import make_cylinder_bell_funnel
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
# warnings.filterwarnings("ignore")
random_state = 0
dataset_name = "cbf"
X, y = make_cylinder_bell_funnel(n_samples=100, random_state=random_state)
print("DATASET INFO:")
print("X SHAPE: ", X.shape)
print("y SHAPE: ", y.shape)
unique, counts = np.unique(y, return_counts=True)
print("\nCLASSES BALANCE")
for i, label in enumerate(unique):
print(label, ": ", round(counts[i] / sum(counts), 2))
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.3, stratify=y, random_state=random_state)
print("\nSHAPES:")
print("TRAINING SET: ", X_train.shape)
print("TEST SET: ", X_test.shape)
def build_cbf(n_samples=600, random_state=0, verbose=True):
X_all, y_all = make_cylinder_bell_funnel(n_samples=n_samples, random_state=random_state)
X_all = X_all[:, :, np.newaxis]
if verbose:
print("DATASET INFO:")
print("X SHAPE: ", X_all.shape)
print("y SHAPE: ", y_all.shape)
unique, counts = np.unique(y_all, return_counts=True)
print("\nCLASSES BALANCE")
for i, label in enumerate(unique):
print(label, ": ", round(counts[i] / sum(counts), 2))
# BLACKBOX/EXPLANATION SETS SPLIT
X_train, X_exp, y_train, y_exp = train_test_split(
X_all,
y_all,
test_size=0.3,
stratify=y_all, random_state=random_state
)
# BLACKBOX TRAIN/TEST SETS SPLIT
X_train, X_test, y_train, y_test = train_test_split(
X_train,
y_train,
test_size=0.2,
stratify=y_train,
random_state=random_state
)
# BLACKBOX TRAIN/VALIDATION SETS SPLIT
X_train, X_val, y_train, y_val = train_test_split(
X_train,
y_train,
test_size=0.2,
stratify=y_train,
random_state=random_state
)
# EXPLANATION TRAIN/TEST SETS SPLIT
X_exp_train, X_exp_test, y_exp_train, y_exp_test = train_test_split(
X_exp,
y_exp,
test_size=0.2,
stratify=y_exp,
random_state=random_state
)
# EXPLANATION TRAIN/VALIDATION SETS SPLIT
X_exp_train, X_exp_val, y_exp_train, y_exp_val = train_test_split(
X_exp_train, y_exp_train,
test_size=0.2,
stratify=y_exp_train,
random_state=random_state
)
if verbose:
print("\nSHAPES:")
print("BLACKBOX TRAINING SET: ", X_train.shape)
print("BLACKBOX VALIDATION SET: ", X_val.shape)
print("BLACKBOX TEST SET: ", X_test.shape)
print("EXPLANATION TRAINING SET: ", X_exp_train.shape)
print("EXPLANATION VALIDATION SET: ", X_exp_val.shape)
print("EXPLANATION TEST SET: ", X_exp_test.shape)
return (X_train, y_train, X_val, y_val, X_test, y_test, X_exp_train,
y_exp_train, X_exp_val, y_exp_val, X_exp_test, y_exp_test)
def build_multivariate_cbf(n_samples=600, n_features=3, random_state=0, verbose=True):
X_all = [[], [], []]
y_all = []
for i in range(n_features):
X, y = make_cylinder_bell_funnel(n_samples=n_samples, random_state=random_state + i)
X = X[:, :, np.newaxis]
for label in range(3):
X_all[label].append(X[np.nonzero(y == label)])
for i in range(len(X_all)):
X_all[i] = np.concatenate(X_all[i], axis=2)
for label in range(3):
y_all.extend(label for i in range(len(X_all[label])))
X_all = np.concatenate(X_all, axis=0)
y_all = np.array(y_all)
if verbose:
print("DATASET INFO:")
print("X SHAPE: ", X_all.shape)
print("y SHAPE: ", y_all.shape)
unique, counts = np.unique(y_all, return_counts=True)
print("\nCLASSES BALANCE")
for i, label in enumerate(unique):
print(label, ": ", round(counts[i] / sum(counts), 2))
# BLACKBOX/EXPLANATION SETS SPLIT
X_train, X_exp, y_train, y_exp = train_test_split(
X_all,
y_all,
test_size=0.3,
stratify=y_all, random_state=random_state
)
# BLACKBOX TRAIN/TEST SETS SPLIT
X_train, X_test, y_train, y_test = train_test_split(
X_train,
y_train,
test_size=0.2,
stratify=y_train,
random_state=random_state
)
# BLACKBOX TRAIN/VALIDATION SETS SPLIT
X_train, X_val, y_train, y_val = train_test_split(
X_train,
y_train,
test_size=0.2,
stratify=y_train,
random_state=random_state
)
# EXPLANATION TRAIN/TEST SETS SPLIT
X_exp_train, X_exp_test, y_exp_train, y_exp_test = train_test_split(
X_exp,
y_exp,
test_size=0.2,
stratify=y_exp,
random_state=random_state
)
# EXPLANATION TRAIN/VALIDATION SETS SPLIT
X_exp_train, X_exp_val, y_exp_train, y_exp_val = train_test_split(
X_exp_train, y_exp_train,
test_size=0.2,
stratify=y_exp_train,
random_state=random_state
)
if verbose:
print("\nSHAPES:")
print("BLACKBOX TRAINING SET: ", X_train.shape)
print("BLACKBOX VALIDATION SET: ", X_val.shape)
print("BLACKBOX TEST SET: ", X_test.shape)
print("EXPLANATION TRAINING SET: ", X_exp_train.shape)
print("EXPLANATION VALIDATION SET: ", X_exp_val.shape)
print("EXPLANATION TEST SET: ", X_exp_test.shape)
return (X_train, y_train, X_val, y_val, X_test, y_test, X_exp_train,
y_exp_train, X_exp_val, y_exp_val, X_exp_test, y_exp_test)