def test_r2(self):
""" Test case for testing the R2 score method. """
#1.)
with warnings.catch_warnings(
record=True) as w: #catch Divide by zero runtime warning
warnings.simplefilter("always")
eval = Evaluate(self.a, self.a)
self.assertEqual(len(w), 1)
self.assertTrue("divide by zero encountered in double_scalars" in
str(w[-1].message))
#assert that R2 = 1 if Y_true and Y_pred arrays are equal
self.assertEqual(eval.r2, 1)
#assert r2 variable is of type float
self.assertIsInstance(eval.r2, float)
#2.)
#############################################
eval = Evaluate(self.c, self.d)
#assert r2 value is around the pre-calculated value for c=Y_true, d=Y_pred
self.assertAlmostEqual(eval.r2, -3.0)
#assert r2 variable is of type float
self.assertIsInstance(eval.r2, float)
#3.)
#############################################
eval = Evaluate(self.e, self.f)
actual_value = r2_score(self.e, self.f)
self.assertAlmostEqual(eval.r2, actual_value)
#assert r2 variable is of type float
self.assertIsInstance(eval.r2, float)
def test_evaluate(self):
""" Test Evaluate class initialisation and its attributes. """
#1.)
#assert that value error is raised when input parameters to class are not same shape
with self.assertRaises(
ValueError,
msg=
'Value Error raised, input parameters are not of the same shape.'
):
fail_eval = Evaluate(self.a, self.b)
eval = Evaluate(self.a, self.a_)
metrics = ['r2', 'mse', 'rmse', 'mae', 'rpd', 'explained_var']
#2.)
#assert metric names exist as instance variables in Evaluate class
for met in range(0, len(metrics)):
self.assertTrue(metrics[met] in list(eval.__dict__.keys()))
#3.)
#testing Y_true variable in the Evaluate object has been set according to Y_true input parameter
np_equality = np.testing.assert_array_equal(self.a, eval.Y_true)
self.assertIsNone(np_equality)
#4.)
#testing Y_pred variable in the Evaluate object has been set according to Y_pred input parameter
np_equality = np.testing.assert_array_equal(self.a_, eval.Y_pred)
self.assertIsNone(np_equality)
#5.)
#numpy.testing function will raise an assertion error if the arrays are not equal
with self.assertRaises(
AssertionError,
msg='Assertion Error raised, input arrays are not equal.'):
np.testing.assert_array_equal(self.a_, eval.Y_true)
#6.)
self.assertTrue(eval.Y_true.shape == (100, 1))
self.assertTrue(eval.Y_pred.shape == (100, 1))
#############################################
eval = Evaluate(self.c, self.d)
metrics = ['r2', 'mse', 'rmse', 'mae', 'rpd', 'explained_var']
#7.)
#assert metric names exist as instance variables in Evaluate class
for met in range(0, len(metrics)):
self.assertTrue(metrics[met] in list(eval.__dict__.keys()))
#8.)
np_equality = np.testing.assert_array_equal(self.c, eval.Y_true)
self.assertIsNone(np_equality)
#9.)
np_equality = np.testing.assert_array_equal(self.d, eval.Y_pred)
self.assertIsNone(np_equality)
#10.)
#numpy.testing function will raise an assertion error if the arrays are not equal
with self.assertRaises(
AssertionError,
msg='Assertion Error raised, input arrays are not equal.'):
np.testing.assert_array_equal(self.d, eval.Y_true)
#11.)
self.assertTrue(eval.Y_true.shape == (10, 1))
self.assertTrue(eval.Y_pred.shape == (10, 1))
def test_mse(self):
""" Test case for testing the MSE method. """
#1.)
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
eval = Evaluate(self.a, self.a)
self.assertEqual(len(w), 1)
self.assertTrue("divide by zero encountered in double_scalars" in
str(w[-1].message))
self.assertEqual(eval.mse, 0)
self.assertIsInstance(eval.mse, float)
#2.)
#############################################
eval = Evaluate(self.c, self.d)
self.assertIsInstance(eval.mse, float)
#3.)
#############################################
eval = Evaluate(self.e, self.f)
self.assertIsInstance(eval.mse, float)
self.assertTrue(eval.mse > 0)
def test_rpd(self):
""" Testing RPD evaluation metric. """
#1.)
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
eval = Evaluate(self.a, self.a)
self.assertEqual(len(w), 1)
self.assertTrue("divide by zero encountered in double_scalars" in
str(w[-1].message))
self.assertEqual(eval.rpd, 0)
self.assertIsInstance(eval.rpd, float)
#2.)
#############################################
eval = Evaluate(self.c, self.d)
self.assertIsInstance(eval.rpd, float)
self.assertTrue(eval.rpd > 0)
#3.)
#############################################
eval = Evaluate(self.e, self.f)
self.assertIsInstance(eval.rpd, float)
self.assertTrue(eval.rpd > 0)
def test_explainedVar(self):
""" Testing Explained Variance metric. """
#1.)
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
eval = Evaluate(self.a, self.a)
self.assertEqual(len(w), 1)
self.assertTrue("divide by zero encountered in double_scalars" in
str(w[-1].message))
self.assertEqual(eval.explained_var, 1)
self.assertIsInstance(eval.explained_var, float)
#2.)
#############################################
eval = Evaluate(self.c, self.d)
self.assertIsInstance(eval.explained_var, float)
self.assertTrue(eval.explained_var == -3)
#3.)
#############################################
eval = Evaluate(self.e, self.f)
self.assertIsInstance(eval.explained_var, float)
self.assertTrue(eval.explained_var <= 1)