def test_mean_averaging():
rng = np.random.RandomState(1)
X = generate_univaritate_series(n=100, size=5, rng=rng, dtype=np.int)
mean = MeanAveraging(X)
average = mean.average()
assert np.array_equal(np.array([-0.02, 0.01, 0.05, 0.11, 0.03]), average)
def test_barycenter_averaging():
"""Test barycenter averaging."""
rng = np.random.RandomState(0)
X = generate_univaritate_series(n=100, size=5, rng=rng, dtype=int)
BCA = BarycenterAveraging(X)
BCA.average()
def test_random_cluster_center_initializer():
n_clusters = 3
k_medians = TimeSeriesKMedoids(n_clusters=n_clusters)
rng = np.random.RandomState(0)
X = generate_univaritate_series(n=20,
size=n_clusters,
rng=rng,
dtype=np.int64)
random_centers_medians = RandomCenterInitializer(
X, n_clusters, k_medians.calculate_new_centers, rng)
centers = random_centers_medians.initialize_centers()
assert np.array_equal(
np.array([[472, 600, 396], [544, 543, 714], [684, 559, 629]]), centers)
k_means = TimeSeriesKMeans(n_clusters=n_clusters)
random_centers_mean = RandomCenterInitializer(
X, n_clusters, k_means.calculate_new_centers, rng)
centers = random_centers_mean.initialize_centers()
assert np.array_equal(
np.array([[520, 521, 692], [491, 581, 409], [695, 492, 403]]), centers)
k_means_dtw = TimeSeriesKMeans(n_clusters=n_clusters, metric="dtw")
random_centers_mean = RandomCenterInitializer(
X, n_clusters, k_means_dtw.calculate_new_centers, rng)
centers = random_centers_mean.initialize_centers()
assert np.array_equal(
np.array([[516, 564, 670], [564, 546, 480], [632, 438, 527]]), centers)
def make_clustering_problem(n_instances=20,
series_size=20,
return_numpy=True,
random_state=None):
# Can only currently support univariate so converting
# to univaritate for the time being
X = generate_univaritate_series(n_instances, series_size, random_state)
if return_numpy:
return X
else:
return pd.Series(X)
def test_k_medoids():
rng = np.random.RandomState(1)
X_train = generate_univaritate_series(n=100,
size=5,
rng=rng,
dtype=np.double)
X_test = generate_univaritate_series(n=10,
size=5,
rng=np.random.RandomState(2),
dtype=np.double)
clusters, _ = run_clustering_experiment(
TimeSeriesKMedoids(n_clusters=5, max_iter=50, random_state=rng),
X_train, X_test)
assert np.array_equal(np.array([1, 1, 0, 0, 0, 0, 1, 1, 1, 0]), clusters)
clusters, _ = run_clustering_experiment(
TimeSeriesKMedoids(n_clusters=5,
max_iter=50,
metric="euclidean",
random_state=rng),
X_train,
X_test,
)
assert np.array_equal(np.array([1, 3, 3, 4, 4, 3, 4, 3, 2, 4]), clusters)
clusters, _ = run_clustering_experiment(
TimeSeriesKMedoids(
n_clusters=5,
max_iter=50,
init_algorithm="random",
metric="euclidean",
random_state=rng,
),
X_train,
X_test,
)
assert np.array_equal(np.array([2, 1, 1, 3, 3, 0, 1, 0, 1, 3]), clusters)
def test_forgy_cluster_center_initializer():
rng = np.random.RandomState(0)
X = generate_univaritate_series(n=20, size=1, rng=rng, dtype=np.float32)
forgy_centers = ForgyCenterInitializer(X, 5, random_state=rng)
centers = forgy_centers.initialize_centers()
assert np.array_equal(
np.array(
[[0.33367434], [1.4940791], [0.95008844], [0.4001572],
[-0.10321885]],
dtype=np.float32,
),
centers,
)
def test_k_means():
rng = np.random.RandomState(1)
X_train = generate_univaritate_series(n=100,
size=5,
rng=rng,
dtype=np.double)
X_test = generate_univaritate_series(n=10,
size=5,
rng=np.random.RandomState(2),
dtype=np.double)
clusters, _ = run_clustering_experiment(
TimeSeriesKMeans(
n_clusters=5,
max_iter=50,
metric="euclidean",
averaging_algorithm="mean",
init_algorithm="forgy",
random_state=rng,
),
X_train,
X_test,
)
assert np.array_equal(np.array([3, 1, 0, 2, 0, 1, 1, 1, 1, 0]), clusters)
# Bug with dtw as metric that is only works if the array is type double
clusters, _ = run_clustering_experiment(
TimeSeriesKMeans(
n_clusters=5,
max_iter=50,
metric="dtw",
averaging_algorithm="mean",
random_state=rng,
),
X_train,
X_test,
)
assert np.array_equal(np.array([2, 3, 3, 4, 4, 3, 3, 3, 2, 3]), clusters)
clusters, _ = run_clustering_experiment(
TimeSeriesKMeans(
n_clusters=5,
max_iter=5,
averaging_algorithm="dba",
averaging_algorithm_iterations=2,
random_state=rng,
),
X_train,
X_test,
)
# Need to add seeding to dba so this works
assert clusters
clusters, _ = run_clustering_experiment(
TimeSeriesKMeans(
n_clusters=5,
max_iter=5,
init_algorithm="random",
averaging_algorithm="dba",
averaging_algorithm_iterations=2,
random_state=rng,
),
X_train,
X_test,
)
assert np.array_equal(np.array([1, 4, 3, 3, 4, 4, 1, 1, 1, 4]), clusters)