def test_dataset_histogram(self) -> None:
# generates 2 ** N - 1 timeseries with constant increasing values
N = 6
n = 2**N - 1
T = 5
targets = np.ones((n, T))
for i in range(0, n):
targets[i, :] = targets[i, :] * i
# FIXME: the cast below is a hack to make mypy happy
timeseries = cast(
Dataset,
[make_time_series(target=targets[i, :]) for i in range(n)])
found = calculate_dataset_statistics(timeseries)
hist = found.scale_histogram.bin_counts
for i in range(0, N):
assert i in hist
assert hist[i] == 2**i
def test_dataset_statistics(self) -> None:
n = 2
T = 10
# use integers to avoid float conversion that can fail comparison
np.random.seed(0)
targets = np.random.randint(0, 10, (n, T))
scale_histogram = ScaleHistogram()
for i in range(n):
scale_histogram.add(targets[i, :])
scale_histogram.add([])
expected = DatasetStatistics(
integer_dataset=True,
num_time_series=n + 1,
num_time_observations=targets.size,
mean_target_length=T * 2 / 3,
min_target=targets.min(),
mean_target=targets.mean(),
mean_abs_target=targets.mean(),
max_target=targets.max(),
feat_static_real=[{0.1}, {0.2, 0.3}],
feat_static_cat=[{1}, {2, 3}],
num_feat_dynamic_real=2,
num_feat_dynamic_cat=2,
num_missing_values=0,
scale_histogram=scale_histogram,
)
# FIXME: the cast below is a hack to make mypy happy
timeseries = cast(
Dataset,
[
make_time_series(
target=targets[0, :],
feat_static_cat=[1, 2],
feat_static_real=[0.1, 0.2],
num_feat_dynamic_cat=2,
num_feat_dynamic_real=2,
),
make_time_series(
target=targets[1, :],
feat_static_cat=[1, 3],
feat_static_real=[0.1, 0.3],
num_feat_dynamic_cat=2,
num_feat_dynamic_real=2,
),
make_time_series(
target=np.array([]),
feat_static_cat=[1, 3],
feat_static_real=[0.1, 0.3],
num_feat_dynamic_cat=2,
num_feat_dynamic_real=2,
),
],
)
found = calculate_dataset_statistics(timeseries)
assert expected == found
def check_error_message(expected_regex, dataset) -> None:
with self.assertRaisesRegex(Exception, expected_regex):
calculate_dataset_statistics(dataset)
def test_dataset_statistics_exceptions(self) -> None:
def check_error_message(expected_regex, dataset) -> None:
with self.assertRaisesRegex(Exception, expected_regex):
calculate_dataset_statistics(dataset)
check_error_message("Time series dataset is empty!", [])
check_error_message(
"Only empty time series found in the dataset!",
[make_time_series(target=np.random.randint(0, 10, 0))],
)
# infinite target
# check_error_message(
# "Target values have to be finite (e.g., not inf, -inf, "
# "or None) and cannot exceed single precision floating "
# "point range.",
# [make_time_series(target=np.full(20, np.inf))]
# )
# different number of feat_dynamic_{cat, real}
check_error_message(
"Found instances with different number of features in "
"feat_dynamic_cat, found one with 2 and another with 1.",
[
make_time_series(num_feat_dynamic_cat=2),
make_time_series(num_feat_dynamic_cat=1),
],
)
check_error_message(
"Found instances with different number of features in "
"feat_dynamic_cat, found one with 0 and another with 1.",
[
make_time_series(num_feat_dynamic_cat=0),
make_time_series(num_feat_dynamic_cat=1),
],
)
check_error_message(
"feat_dynamic_cat was found for some instances but not others.",
[
make_time_series(num_feat_dynamic_cat=1),
make_time_series(num_feat_dynamic_cat=0),
],
)
check_error_message(
"Found instances with different number of features in "
"feat_dynamic_real, found one with 2 and another with 1.",
[
make_time_series(num_feat_dynamic_real=2),
make_time_series(num_feat_dynamic_real=1),
],
)
check_error_message(
"Found instances with different number of features in "
"feat_dynamic_real, found one with 0 and another with 1.",
[
make_time_series(num_feat_dynamic_real=0),
make_time_series(num_feat_dynamic_real=1),
],
)
check_error_message(
"feat_dynamic_real was found for some instances but not others.",
[
make_time_series(num_feat_dynamic_real=1),
make_time_series(num_feat_dynamic_real=0),
],
)
# infinite feat_dynamic_{cat,real}
inf_dynamic_feat = np.full((2, len(target)), np.inf)
check_error_message(
"Features values have to be finite and cannot exceed single "
"precision floating point range.",
[
ts(
start,
target,
feat_dynamic_cat=inf_dynamic_feat,
feat_static_cat=[0, 1],
)
],
)
check_error_message(
"Features values have to be finite and cannot exceed single "
"precision floating point range.",
[
ts(
start,
target,
feat_dynamic_real=inf_dynamic_feat,
feat_static_cat=[0, 1],
)
],
)
# feat_dynamic_{cat, real} different length from target
check_error_message(
"Each feature in feat_dynamic_cat has to have the same length as the "
"target. Found an instance with feat_dynamic_cat of length 1 and a "
"target of length 20.",
[
ts(
start=start,
target=target,
feat_static_cat=[0, 1],
feat_dynamic_cat=np.ones((1, 1)),
)
],
)
check_error_message(
"Each feature in feat_dynamic_real has to have the same length as the "
"target. Found an instance with feat_dynamic_real of length 1 and a "
"target of length 20.",
[
ts(
start=start,
target=target,
feat_static_cat=[0, 1],
feat_dynamic_real=np.ones((1, 1)),
)
],
)
# feat_static_{cat, real} different length
check_error_message(
"Not all feat_static_cat vectors have the same length 2 != 1.",
[
ts(start=start, target=target, feat_static_cat=[0, 1]),
ts(start=start, target=target, feat_static_cat=[1]),
],
)
check_error_message(
"Not all feat_static_real vectors have the same length 2 != 1.",
[
ts(start=start, target=target, feat_static_real=[0, 1]),
ts(start=start, target=target, feat_static_real=[1]),
],
)
calculate_dataset_statistics(
# FIXME: the cast below is a hack to make mypy happy
cast(
Dataset,
[
make_time_series(num_feat_dynamic_cat=2),
make_time_series(num_feat_dynamic_cat=2),
],
))
calculate_dataset_statistics(
# FIXME: the cast below is a hack to make mypy happy
cast(
Dataset,
[
make_time_series(num_feat_dynamic_cat=0),
make_time_series(num_feat_dynamic_cat=0),
],
))