def seasonal_features(cls, freq: str) -> List[TimeFeature]:
offset = to_offset(freq)
if offset.name == "M":
return [MonthOfYearIndex()]
elif offset.name == "W-SUN":
return [WeekOfYearIndex()]
elif offset.name == "D":
return [DayOfWeekIndex()]
elif offset.name == "B": # TODO: check this case
return [DayOfWeekIndex()]
elif offset.name == "H":
return [HourOfDayIndex(), DayOfWeekIndex()]
elif offset.name == "T":
return [
MinuteOfHourIndex(),
HourOfDayIndex(),
]
else:
RuntimeError(f"Unsupported frequency {offset.name}")
return []
DayOfWeekIndex(),
pd.date_range("01-01-2015", periods=365 * 5, freq="D"),
7,
),
(
DayOfMonthIndex(),
pd.date_range("01-01-2015", periods=365 * 5, freq="D"),
31,
),
(
DayOfYearIndex(),
pd.date_range("01-01-2015", periods=365 * 5, freq="D"),
366,
),
(
WeekOfYearIndex(),
pd.date_range("01-01-2015", periods=53 * 5, freq="W"),
53,
),
(
MonthOfYearIndex(),
pd.date_range("01-01-2015", periods=12 * 5, freq="M"),
12,
),
],
)
def test_feature_unnormalized_bounds(feature: TimeFeature,
index: pd.DatetimeIndex,
cardinality: int):
values = feature(index)
assert isinstance(values, np.ndarray)
def WeekOfYearSeasonalISSM():
return SeasonalityISSM(num_seasons=53, time_feature=WeekOfYearIndex())