def test_result_format_fit(self):
# Load Test Data
data = TestData()
X, y = data.make_X_y()
robo = RoboLogistic(random_state=0)
# Fit returns None
res_fit = robo.fit(X, y)
self.assertIsNone(res_fit)
def test_result_format_predict(self):
# Load Test Data
data = TestData()
X, y = data.make_X_y()
robo = RoboLogistic(random_state=0)
robo.fit(X, y)
# Output Type np.ndarray
# Output Example np.array([1, 0, 1])
res_predict = robo.predict(X)
self.assertIsInstance(res_predict, np.ndarray)
self.assertTupleEqual(res_predict.shape, (data.df.shape[0], ))
def test_bad_input_shape(self):
# Check if transform method input shape mismatches fit method input
# and it will raise ValueError("Shape of input ...")
# Load Test Data
data = TestData()
X, y = data.make_X_y()
# Fit
robo = RoboLogistic()
robo.fit(X, y)
# Only Columns, shape[1] should be the same
self.assertRaises(ValueError, robo.predict, X.iloc[:, ::2])
self.assertRaises(ValueError, robo.predict_proba, X.iloc[:, ::2])
def test_result_format_predict_proba(self):
# Load Test Data
data = TestData()
X, y = data.make_X_y()
robo = RoboLogistic(random_state=0)
robo.fit(X, y)
# Output Type np.ndarray
# Output Example np.array([[0.2, 0.8], [0.9, 0.1], [0.5, 0.5]])
res_predict_proba = robo.predict_proba(X)
self.assertIsInstance(res_predict_proba, np.ndarray)
self.assertEqual(res_predict_proba.shape,
(data.df.shape[0], len(robo.classes_)))
self.assertTupleEqual(res_predict_proba.shape,
(data.df.shape[0], len(robo.classes_)))
def test_result_format_tune(self):
# Load Test Data
data = TestData()
X, y = data.make_X_y()
robo = RoboLogistic(random_state=0)
# Output Type dict
# Output Example {'tol': 0.02, 'fit_intercept': False, 'solver': 'sag', ‘scores’:
# {'f1_score': 0.3, 'logloss': 0.7}}
res_tune_parameters = robo.tune_parameters(X, y)
self.assertIsInstance(res_tune_parameters, dict)
self.assertTrue('scores' in res_tune_parameters.keys())
self.assertTrue('f1_score' in res_tune_parameters['scores'].keys())
self.assertTrue('logloss' in res_tune_parameters['scores'].keys())
def test_result_format_evaluate(self):
# Load Test Data
data = TestData()
X, y = data.make_X_y()
robo = RoboLogistic(random_state=0)
robo.fit(X, y)
# Output Type dict
# Output Example {'f1_score': 0.3, 'logloss': 0.7}
res_evaluate = robo.evaluate(X, y)
self.assertIsInstance(res_evaluate, dict)
self.assertTrue('f1_score' in res_evaluate.keys())
self.assertTrue('logloss' in res_evaluate.keys())
self.assertIsInstance(res_evaluate['f1_score'], np.float)
self.assertIsInstance(res_evaluate['logloss'], np.float)
def test_reproduce(self):
# Load Test Data
data = TestData()
X, y = data.make_X_y()
robo = RoboLogistic(random_state=10)
robo.fit(X, y)
res_evaluate = robo.evaluate(X, y)
if not data.skcancer:
self.assertDictEqual(res_evaluate, {
'f1_score': 0.003084040092521203,
'logloss': 0.37192273211820986
})
else:
self.assertDictEqual(res_evaluate, {
'f1_score': 0.9902370990237099,
'logloss': 0.05331853517496121
})
def test_unfit_instance(self):
# Check if unfitted instance is called,
# then it will raise NotFittedError!
# Load Test Data
data = TestData()
X, y = data.make_X_y()
robo = RoboLogistic()
self.assertRaises(NotFittedError, robo.predict, X)
self.assertRaises(NotFittedError, robo.predict_proba, X)
self.assertRaises(NotFittedError, robo.evaluate, X, y)