def test_get_labels(self):
a = EncodedAngle(segments=2)
X = a.fit_transform([METHANE])
labels = a.get_labels()
self.assertEqual(X.shape[1], len(labels))
expected = (
'C-H-C_0.0', 'C-H-C_3.14159',
'C-H-H_0.0', 'C-H-H_3.14159',
'H-C-H_0.0', 'H-C-H_3.14159',
'H-H-H_0.0', 'H-H-H_3.14159',
)
self.assertEqual(labels, expected)
def test_fit_transform(self):
a = EncodedAngle()
# This is a cheap test to prevent needing all the values here
expected_results = numpy.array([
0.116708, # mean
0.450738, # std
0., # min
3.043729, # max
])
try:
m = a.fit_transform([METHANE])
assert_close_statistics(m, expected_results)
except AssertionError as e:
self.fail(e)
def test_smoothing_function(self):
a = EncodedAngle(smoothing="norm_cdf")
# This is a cheap test to prevent needing all the values here
expected_results = numpy.array([
1.5891686, # mean
2.5907034, # std
0., # min
9.8982443, # max
])
try:
m = a.fit_transform([METHANE])
assert_close_statistics(m, expected_results)
except AssertionError as e:
self.fail(e)
def test_small_to_large_transform(self):
a = EncodedAngle()
a.fit([METHANE])
# This is a cheap test to prevent needing all the values here
expected_results = numpy.array([
0.018603, # mean
0.130329, # std
0., # min
1.568823, # max
])
try:
m = a.transform([MID])
assert_close_statistics(m, expected_results)
except AssertionError as e:
self.fail(e)
def test_form(self):
data = (
# mean std min max
(2, [0.155611, 0.581838, 0., 4.395692], 120),
(1, [0.233417, 0.699744, 0., 4.395692], 80),
(0, [4.668338e-001, 1.090704e+000, 0., 5.747656e+000], 40),
)
for form, expected, size in data:
a = EncodedAngle(form=form)
expected_results = numpy.array(expected)
try:
m = a.fit_transform([METHANE])
self.assertEqual(m.shape, (1, size))
assert_close_statistics(m, expected_results)
except AssertionError as e:
self.fail(e)
def test_add_unknown(self):
a = EncodedAngle(add_unknown=True)
a.fit([METHANE])
# This is a cheap test to prevent needing all the values here
expected_results = numpy.array([
0.117057, # mean
0.510819, # std
0., # min
6.343512, # max
])
try:
m = a.transform([MID])
self.assertEqual(m.shape, (1, 200))
assert_close_statistics(m, expected_results)
except AssertionError as e:
self.fail(e)
def test_max_depth(self):
a = EncodedAngle(max_depth=3)
# This is a cheap test to prevent needing all the values here
data = (
# mean std min max
(-1, [0.0325158765862, 0.132101907024, 0.0, 2.01566683797]),
(1, [0.00491078348799, 0.0463273875823, 0.0, 0.694568644823]),
(3, [0.0063668265711, 0.0513782485995, 0.0, 0.694568644823]),
)
for max_depth, expected in data:
a = EncodedAngle(max_depth=max_depth)
expected_results = numpy.array(expected)
try:
m = a.fit_transform([MID])
assert_close_statistics(m, expected_results)
except AssertionError as e:
self.fail(e)
def test_fit(self):
a = EncodedAngle()
a.fit(ALL_DATA)
expected = set([('C', 'N', 'C'), ('C', 'C', 'C'), ('H', 'H', 'H'),
('H', 'O', 'O'), ('O', 'N', 'O'), ('H', 'N', 'N'),
('C', 'H', 'H'), ('C', 'O', 'H'), ('C', 'H', 'C'),
('N', 'C', 'N'), ('O', 'O', 'O'), ('H', 'O', 'N'),
('H', 'N', 'O'), ('O', 'H', 'O'), ('H', 'H', 'N'),
('C', 'C', 'N'), ('H', 'N', 'H'), ('C', 'H', 'N'),
('H', 'C', 'O'), ('N', 'O', 'O'), ('N', 'N', 'N'),
('C', 'C', 'H'), ('C', 'O', 'O'), ('C', 'N', 'N'),
('H', 'O', 'H'), ('H', 'H', 'O'), ('C', 'C', 'O'),
('N', 'H', 'N'), ('C', 'H', 'O'), ('O', 'C', 'O'),
('H', 'C', 'N'), ('C', 'O', 'C'), ('N', 'O', 'N'),
('N', 'N', 'O'), ('C', 'N', 'O'), ('C', 'O', 'N'),
('H', 'C', 'H'), ('C', 'N', 'H'), ('N', 'H', 'O'),
('N', 'C', 'O')])
self.assertEqual(a._groups, expected)
def test_smoothing_function_error(self):
a = EncodedAngle(smoothing="not valid")
with self.assertRaises(KeyError):
a.fit_transform([METHANE])
def test_transform_before_fit(self):
a = EncodedAngle()
with self.assertRaises(ValueError):
a.transform(ALL_DATA)