def test_shap(self) -> None:
from algoneer.methods.blackbox.shap import SHAP, SHAPDatapointResult, SHAPModelResult
project = Project("test")
# we load the dataset
dataset = load_dataset(project)
# we select an algorithm from the set of algorithms
algo = get_algorithm(project, 'linear-regression')
# we train the algorithm with the dataset to obtain a model
model = algo.fit(dataset)
# we initialize a SHAP test
shap = SHAP()
# we run the SHAP test on the model and dataset
result = shap.run(dataset, model, max_datapoints=10)
# we make sure that we obtain a reasonable result
assert isinstance(result, DatasetModelResult)
assert 'expected_value' in result.data
def test_ale(self) -> None:
project = Project("test")
# we load the dataset
dataset = load_dataset(project)
# we select an algorithm from the set of algorithms
algo = get_algorithm(project, 'linear-regression')
# we train the algorithm with the dataset to obtain a model
model = algo.fit(dataset)
# we initialize a ALE test
ale = ALE()
columns = ['windspeed', 'hum', 'atemp', 'season']
# we run the ALE test on the model and dataset
result = ale.run(dataset,
model,
columns=columns,
max_datapoints=200,
n_intervals=10)
# we make sure that we obtain a reasonable result
assert isinstance(result, DatasetModelResult)
def test_sklean(self):
project = Project("test")
# we load the dataset
dataset = load_dataset(project)
assert isinstance(dataset, Dataset)
hash = dataset.hash
assert hash == b'\x01\x00\x00\x00\xb9\x96\xdb_\xdc\xe6\x98\x05\xe0\xe5\xec`HQ\x82\x14\xde[o5\n\xff\xd2\xb1\xe2K\x06>\xd1:M\xf9'
datapoint = dataset.datapoint(0)
dp_hash = datapoint.hash
assert dp_hash == b'\x01\x00\x00\x00E/\xbdy\xbe#\x19g\xf3\xafV\x1a\xe3\x84\xeel{:\xde)\xa0\x89e\xc1\xbf\xd9\xc6\x94\x8a\x90G\xde'
def test_sklean(self):
from algoneer.algorithm.sklearn import SklearnAlgorithm
from sklearn.ensemble import RandomForestRegressor
project = Project("test")
# we load the dataset
dataset = load_dataset(project)
algo = SklearnAlgorithm(project, RandomForestRegressor, n_estimators=100)
assert isinstance(algo, Algorithm)
assert 'kwargs' in algo.data and algo.data['kwargs'] == {"n_estimators" : 100, "random_state" : 0}
model = algo.fit(dataset)
assert isinstance(model, Model)
def test_objects(self) -> None:
project = Project("test")
# we load the dataset
dataset = load_dataset(project)
# we select an algorithm from the set of algorithms
algo = get_algorithm(project, 'linear-regression')
# we train the algorithm with the dataset to obtain a model
model = algo.fit(dataset)
assert isinstance(algo, Object)
assert isinstance(model, Object)
assert isinstance(dataset, Object)
client: Union[Client, TestClient]
token = os.environ.get("ALGONAUT_TOKEN")
if token:
client = Client(token)
else:
client = TestClient()
session = Session(client)
for Test in dataset_model_tests:
# we initialize the test
test = Test()
# we run the test on the model and dataset
result = test.run(dataset, model, max_datapoints=10, max_values=10)
session.add(result)
api_model = session.add(model)
api_dataset = session.add(dataset)
api_algo = session.add(algo)
assert model in session
assert dataset in session
assert algo in session
session.sync()
def test_predictions(self) -> None:
project = Project("test")
# we load the dataset
dataset = load_dataset(project)
# we select an algorithm from the set of algorithms
algo = get_algorithm(project, 'random-forest')
# we train the algorithm with the dataset to obtain a model
model = algo.fit(dataset)
# we initialize a predictions test
predictions = Predictions()
# we run the predictions test on the model and dataset
result = predictions.run(dataset, model, max_datapoints=110)
# we make sure that we obtain a reasonable result
assert isinstance(result, DatasetModelResult)