def test_fit_ko_from_bad_folds(self):
"""
Tests the execution when the user gives bad number of folds
"""
with self.assertRaises(IkatsInputTypeError):
fit(population="id", target_column_name='"Species"',
identifier_column_name="Id", depth_parameters="0;3;5;4",
balanced_parameters="True;False", table_name="my_table", folds=2.4)
def test_fit_ko_from_bad_balancing(self):
"""
Tests the execution when the user gives bad balancing parameters
"""
with self.assertRaises(IkatsException):
fit(population=get_table(), target_column_name='"Species"',
identifier_column_name="Id", depth_parameters="0;3;5;4",
balanced_parameters="true;false", table_name="my_table", folds=3)
def test_fit_population_sklearn_ko(self):
"""
Tests the unexpected error, raised by algorithm
"""
with self.assertRaises(IkatsException):
fit(population=get_table(), target_column_name='"Species"',
identifier_column_name="Id", depth_parameters="0;2;5;3",
table_name="my_table",
balanced_parameters="True;False")
def test_fit_nominal_from_id(self):
"""
Tests the nominal execution based upon mock data: IRIS data
"""
_, mdl, dot, best_params, table_name = fit(population=get_table(), target_column_name='"Species"',
table_name="my_table",
identifier_column_name="Id")
cv_res = IkatsApi.table.read(name=table_name)
# trained: 71,5.9,3.2,4.8,1.8,I. versicolor
# => test below should obtain same class:
#
# mdl.predict([[5.9, 3.2, 4.8, 1.8]]) returns a numpy.ndarray
# so we have to convert predicted value to str:
for value, ref_predict in [[[5.9, 3.2, 4.8, 1.8], "I. versicolor"],
[[5.7, 2.8, 4.1, 1.3, ], "I. versicolor"],
[[5.9, 3.0, 5.1, 1.8], "I. virginica"]]:
predicted = "{}".format(mdl.predict([value])[0])
self.assertEqual(predicted, ref_predict, "Failed to use the computed model: predict x={}".format(value))
self.assertTrue(type(dot) == str, "Bad type for the dot returned")
self.assertTrue(type(best_params) == str, "Bad type for the best parameters returned")
self.assertTrue(type(cv_res) == defaultdict, "Bad type for the cross validation results returned")
def test_fit_from_id_with_param(self):
"""
Tests the nominal execution based upon mock data: IRIS data
"""
_, mdl, dot, best_params, table_name = fit(population=get_table(), target_column_name='"Species"',
identifier_column_name="Id", depth_parameters="0;2;5;3",
balanced_parameters="True;False", table_name="my_table", folds=5)
cv_res = IkatsApi.table.read(name=table_name)
# trained: 71,5.9,3.2,4.8,1.8,I. versicolor
# => test below should obtain same class:
#
# mdl.predict([[5.9, 3.2, 4.8, 1.8]]) returns a numpy.ndarray
# so we have to convert predicted value to str:
predicted = "{}".format(mdl.predict([[5.9, 3.0, 5.1, 1.8]])[0])
self.assertEqual(predicted, "I. virginica", "Failed to use the computed model")
self.assertTrue(type(dot) == str, "Bad type for the dot returned")
self.assertTrue(type(best_params) == str, "Bad type for the best parameters returned")
self.assertTrue(type(cv_res) == defaultdict, "Bad type for the cross validation results returned")