def days_to_first_event(df, groupby, time_col):
""" Calculate days to the first date for each group, in a Time series """
dates = df[time_col].astype("datetime64[ns]").values
ids = df[groupby].values
result = wnp.group_apply(dates, ids, _time_to_min_date)
result = _convert_ns_to_days(result)
return result
def test_vector_group_apply_works_with_2dims():
values = np.array([1, 1, 1, 2, 2, 2])
apply_func = np.sum
ids = np.array([[1, 0], [1, 0], [1, 0], [1, 1], [2, 2], [2, 2]])
expected = np.array([3, 3, 3, 2, 4, 4])
output = utils_np.group_apply(values, ids, apply_func)
assert np.all(output == expected)
def grouped_lagged_decay(df, groupby, col, fillna=0, decay=1):
""" Grouped lagged decay """
values = wnp.fillna(df[col].values, 0)
f = partial(lagged_decay, decay=decay)
result = wnp.group_apply(values, df[groupby].values, f)
result = wnp.fillna(result, fillna)
return result
def grouped_days_since_result(
df, groupby, col="win_flag", value=1, fillna=-1, coldate="scheduled_time"
):
func = partial(days_since_result, value=1)
result = wnp.group_apply(
df[[col, coldate]].values, df[groupby].values, func, multiarg=True
)
result = wnp.fillna(result, fillna)
return result
def test_vector_group_apply():
test_data = [
# values, ids, expected
(
np.array([1, 0, 0, 1, 1, 0]),
np.array([1, 2, 1, 2, 1, 2]),
np.array([1, 0, 1, 1, 2, 1]),
),
(
np.array([1, 0, 0, 1, 1, 0]),
np.array(["A", "B", "A", "B", "A", "B"]),
np.array([1, 0, 1, 1, 2, 1]),
),
]
for values, ids, expected in test_data:
output = utils_np.group_apply(values, ids, np.cumsum)
assert np.all(output == expected)
def test_vector_group_apply_works_with_tuple_ids():
values = np.array([1, 1, 1, 2, 2, 2])
all_ids = [
np.array(
list(
zip(np.array([1, 1, 1, 2, 2, 2]), np.array([0, 0, 0, 1, 2,
2])))),
np.array(
list(
zip(
np.array([1, 1, 1, 2, 2, 2]),
np.array(["A", "A", "A", "B", "C", "C"]),
))),
np.array(
list(
zip(
np.array(["a", "a", "a", "b", "b", "b"]),
np.array(["A", "A", "A", "B", "C", "C"]),
))),
]
expected = np.array([3, 3, 3, 2, 4, 4])
for ids in all_ids:
output = utils_np.group_apply(values, ids, np.sum)
assert np.all(output == expected)
def grouped_ema(df, col, alpha, groupby):
v = df[col].values
func = partial(ema, alpha=alpha)
result = wnp.group_apply(v, df[groupby].values, func)
return result
