def test_generate_models(self):
degs_msg = "generate_models should return one model for each given degree"
list_type_msg = "generate_models should return a list of models"
array_type_msg = (
"each model returned by generate_models should be of type pylab.array"
)
coefficient_mismatch = "coefficients of returned model are not as expected"
# simple y = x case.
x = pylab.array(range(50))
y = pylab.array(range(50))
degrees = [1]
models = ps5.generate_models(x, y, degrees)
self.assertEqual(len(models), len(degrees), degs_msg)
self.assertIsInstance(models, list, list_type_msg)
self.assertIsInstance(models[0], pylab.ndarray, array_type_msg)
self.assertListEqual(list(models[0]), list(pylab.polyfit(x, y, 1)),
coefficient_mismatch)
# two models for y = 2x case
y = pylab.array(range(0, 100, 2))
degrees = [1, 2]
models = ps5.generate_models(x, y, degrees)
self.assertEqual(len(models), len(degrees), degs_msg)
self.assertIsInstance(models, list, list_type_msg)
for m in models:
self.assertIsInstance(m, pylab.ndarray, array_type_msg)
for i in range(2):
self.assertListEqual(
list(models[i]),
list(pylab.polyfit(x, y, degrees[i])),
coefficient_mismatch,
)
# three models
degrees = [1, 2, 20]
models = ps5.generate_models(x, y, degrees)
self.assertEqual(len(models), len(degrees), degs_msg)
self.assertIsInstance(models, list, list_type_msg)
for m in models:
self.assertIsInstance(m, pylab.ndarray, array_type_msg)
for i in range(3):
self.assertListEqual(
list(models[i]),
list(pylab.polyfit(x, y, degrees[i])),
coefficient_mismatch,
)
def test_evaluate_models_on_testing(self):
x = np.array(range(50))
y = np.array(range(0, 100, 2))
degrees = [1, 2]
models = ps5.generate_models(x, y, degrees)
rmse = ps5.evaluate_models_on_testing(x, y, models, False)
correct_rmse = [0, 0]
self.assertEqual(len(rmse), len(correct_rmse),
"RMSE values did not match number of models")
for index in range(len(correct_rmse)):
good_enough = math.isclose(correct_rmse[index], rmse[index])
self.assertTrue(good_enough, "Returned incorrect RMSE values")
def test_evaluate_models_on_training(self):
x = np.array(range(50))
y = np.array(range(0, 100, 2))
degrees = [1, 2]
models = ps5.generate_models(x, y, degrees)
r2 = ps5.evaluate_models_on_training(x, y, models, False)
correct_r2 = [1, 1]
self.assertEqual(len(r2), len(correct_r2),
"Returned incorrect r^2 values")
for index in range(len(correct_r2)):
good_enough = math.isclose(correct_r2[index], r2[index])
self.assertTrue(good_enough, "Returned incorrect r^2 values")