def test_arboreal_set_params(self):
# test that the set_params function sets parameters
at = ArborealTree()
at.set_params(**self.nondefault_vals)
# assert at has the right number of attributes set:
user_attributes = [
a for a in vars(at) if not (a.startswith("_") or a.endswith("_"))
]
self.assertEqual(len(user_attributes), len(self.nondefault_vals))
# now check equality of nondefault_vals and set values:
for arg_name, nondefault_val in self.nondefault_vals.items():
self.assertEqual(getattr(at, arg_name), nondefault_val)
def test_arboreal_constructor(self):
# test that the constructor receives input args and sets them
at = ArborealTree()
# assert at has the right number of attributes set:
user_attributes = [
a for a in vars(at) if not (a.startswith("_") or a.endswith("_"))
]
self.assertEqual(len(user_attributes), len(self.default_vals))
# now check equality of default values and set values:
for arg_name, default_val in self.default_vals.items():
self.assertEqual(getattr(at, arg_name), default_val)
def test_grid_search_interface_for_arboreal_tree(self):
param_grid = {
"bootstrap_criterion_fraction_threshold": [0.8, 0.999_999_999],
"maximum_bootstrap_branching_factor": [1, 2, 3],
"feature_subset_fraction": [None, 0.4],
"min_samples_split": [2, 3],
}
arb = ArborealTree(random_seed=self.random_seed)
gs = GridSearchCV(
arb,
param_grid,
cv=3,
scoring="accuracy",
iid=True,
verbose=0,
n_jobs=self.n_jobs,
) # verbose = 2 for helpful info during the search
gs.fit(self.X_train, self.y_train)
gs_accuracy = gs.score(self.X_test, self.y_test)
expected_accuracy = 0.90
self.assertAlmostEqual(gs_accuracy, expected_accuracy, places=2)
def test_arboreal_set_then_get_params(self):
at = ArborealTree()
at.set_params(**self.nondefault_vals)
got_params = at.get_params()
self.assertEqual(got_params, self.nondefault_vals)
def test_arboreal_get_params(self):
at = ArborealTree()
got_params = at.get_params()
self.assertEqual(got_params, self.default_vals)
def test_arboreal_constructor_and_set_params_are_equal(self):
at_1 = ArborealTree(**self.default_vals)
at_2 = ArborealTree()
at_2.set_params(**self.default_vals)
self.assertEqual(at_1.get_params(), at_2.get_params())
at_3 = ArborealTree(**self.nondefault_vals)
at_4 = ArborealTree()
at_4.set_params(**self.nondefault_vals)
self.assertEqual(at_3.get_params(), at_4.get_params())
"sepal length (cm)",
"sepal width (cm)",
"petal length (cm)",
"petal width (cm)",
]
m.categoricals = ["target"]
m.target = "target"
# Create an Arboreal Dataset for train and test
train_dataset = Dataset(metadata=m, datapoints=train_datapoints)
test_dataset = Dataset(
metadata=m, datapoints=test_datapoints, validate_target=False
)
# Fit an ArborealTree on the train set
tree = ArborealTree() # or DecisionTree() or RandomForest()
tree.fit(train_dataset)
# Predict data points in the test set
predictions = tree.transform(test_dataset)
# Evaluate our performance on the test set
results = []
prediction_datatypes = set()
for dp in test_datapoints_for_eval:
target = dp["target"]
prediction = predictions[dp["identifier"]]
predicted_value = prediction[0]
prediction_datatype = prediction[1]
prediction_datatypes.add(prediction_datatype)
assert (