def test_fit_exception(self, strategy, expectation):
discretizer = KBinsDiscretizer(strategy=strategy)
data = pd.DataFrame({"variable": list(range(0, 10)) + [np.nan]})
with expectation:
discretizer.fit(data, ["variable"])
def test_fit_column(self, n_bins, auto_adapt_bins, data, expected):
discretizer = KBinsDiscretizer(n_bins=n_bins,
auto_adapt_bins=auto_adapt_bins)
actual = discretizer._fit_column(data, column_name="variable")
assert actual == expected
def test_set_attributes_from_dict(self, attribute):
discretizer = KBinsDiscretizer()
params = {
"n_bins": 5,
"strategy": "uniform",
"closed": "left",
"auto_adapt_bins": True,
"starting_precision": 1,
"label_format": "[,)",
"change_endpoint_format": True,
"_bins_by_column": {
"variable": [[0.0, 3.0], [3.0, 6.0], [6.0, 9.0]]
}
}
expected = params[attribute]
if attribute == "_bins_by_column":
# list of list is transformed to a list of tuples
# in KBinsDiscretizer!!!
expected = {"variable": [(0.0, 3.0), (3.0, 6.0), (6.0, 9.0)]}
discretizer.set_attributes_from_dict(params)
actual = getattr(discretizer, attribute)
assert actual == expected
def test_create_bin_labels(self, change_endpoint_format, closed, bins,
expected):
discretizer = KBinsDiscretizer(
closed=closed, change_endpoint_format=change_endpoint_format)
actual = discretizer._create_bin_labels(bins)
assert actual == expected
def test_compute_minimal_precision_of_bin_edges(self, bin_edges,
starting_precision,
expected):
discretizer = KBinsDiscretizer(starting_precision=starting_precision)
actual = discretizer._compute_minimal_precision_of_bin_edges(bin_edges)
assert actual == expected
def test_compute_bin_edges(self, strategy, n_bins, data, expected):
discretizer = KBinsDiscretizer(strategy=strategy)
actual = discretizer._compute_bin_edges(data,
column_name="variable",
n_bins=n_bins,
col_min=data.variable.min(),
col_max=data.variable.max())
assert actual == expected
def test_transform(self, scenario, expectation):
discretizer = KBinsDiscretizer(n_bins=3, strategy="uniform")
data = pd.DataFrame({"variable": ([1] * 10)})
expected = data.copy()
if scenario == "regular_test":
# overwrite data and expected with DataFrame containing
# a non-constant variable
data = pd.DataFrame({"variable": list(range(0, 10)) + [np.nan]})
expected = data.copy()
discretizer.fit(data, ["variable"])
categories = ["0.0 - 3.0", "3.0 - 6.0", "6.0 - 9.0", "Missing"]
expected["variable_bin"] = pd.Categorical(
["0.0 - 3.0"] * 4 + ["3.0 - 6.0"] * 3 + ["6.0 - 9.0"] * 3 +
["Missing"],
categories=categories,
ordered=True)
elif scenario == "constant_data":
discretizer.fit(data, ["variable"])
with expectation:
actual = discretizer.transform(data, ["variable"])
pd.testing.assert_frame_equal(actual, expected)
def test_transform_column(self):
data = pd.DataFrame({"variable": list(range(0, 10)) + [np.nan]})
discretizer = KBinsDiscretizer(n_bins=3, strategy="uniform")
bins = [(0.0, 3.0), (3.0, 6.0), (6.0, 9.0)]
actual = discretizer._transform_column(data, "variable", bins)
categories = ["0.0 - 3.0", "3.0 - 6.0", "6.0 - 9.0", "Missing"]
expected = pd.DataFrame({"variable": list(range(0, 10)) + [np.nan]})
expected["variable_bin"] = pd.Categorical(
["0.0 - 3.0"] * 4 + ["3.0 - 6.0"] * 3 + ["6.0 - 9.0"] * 3 +
["Missing"],
categories=categories,
ordered=True)
# assert using pandas testing module
pd.testing.assert_frame_equal(actual, expected)
def test_attributes_to_dict(self):
discretizer = KBinsDiscretizer()
bins = [(0.0, 3.0), (3.0, 6.0), (6.0, 9.0)]
discretizer._bins_by_column = {"variable": bins}
actual = discretizer.attributes_to_dict()
expected = {
"n_bins": 10,
"strategy": "quantile",
"closed": "right",
"auto_adapt_bins": False,
"starting_precision": 0,
"label_format": "{} - {}",
"change_endpoint_format": False,
"_bins_by_column": {
"variable": [[0.0, 3.0], [3.0, 6.0], [6.0, 9.0]]
}
}
assert actual == expected
def test_compute_bins_from_edges(self, bin_edges, expected):
discretizer = KBinsDiscretizer()
actual = discretizer._compute_bins_from_edges(bin_edges)
assert actual == expected
def test_validate_n_bins_exception(self, n_bins, expectation):
with expectation:
assert KBinsDiscretizer()._validate_n_bins(n_bins=n_bins) is None