def _compare_with_sklearn_mse_regressor_iris(n_estimators=100, sample_weight=None,
description=""):
x_train, x_test, y_train, y_test = \
train_test_split(IRIS.data, IRIS.target,
test_size=0.33, random_state=31)
scikit_model = ScikitRandomForestRegressor(n_estimators=n_estimators,
max_depth=None,
random_state=777)
daal4py_model = DaalRandomForestRegressor(n_estimators=n_estimators,
max_depth=None,
random_state=777)
scikit_predict = scikit_model.fit(
x_train, y_train,
sample_weight=sample_weight).predict(x_test)
daal4py_predict = daal4py_model.fit(
x_train, y_train,
sample_weight=sample_weight).predict(x_test)
scikit_mse = mean_squared_error(scikit_predict, y_test)
daal4py_mse = mean_squared_error(daal4py_predict, y_test)
ratio = daal4py_mse / scikit_mse
reason = description + f"scikit_mse={scikit_mse}, daal4py_mse={daal4py_mse}"
assert ratio <= MSE_RATIO, reason
def _test_mse_regressor_sample_weight(weight):
for _ in range(N_TRIES):
x_train, x_test, y_train, y_test = \
train_test_split(IRIS.data, IRIS.target,
test_size=0.33, random_state=31)
scikit_model = ScikitRandomForestRegressor(n_estimators=100,
max_depth=None,
random_state=777)
daal4py_model = DaalRandomForestRegressor(n_estimators=100,
max_depth=None,
random_state=777)
scikit_predict = scikit_model.fit(
x_train, y_train, sample_weight=weight[0][:100]).predict(x_test)
daal4py_predict = daal4py_model.fit(
x_train, y_train, sample_weight=weight[0][:100]).predict(x_test)
scikit_mse = mean_squared_error(scikit_predict, y_test)
daal4py_mse = mean_squared_error(daal4py_predict, y_test)
ratio = daal4py_mse / scikit_mse
reason = ("Regression sample weights: sample_weight_type={},"
"scikit_mse={}, daal4py_mse={}".format(
weight[1], scikit_mse, daal4py_mse))
assert ratio <= MSE_RATIO, reason
def check_regressor_sample_weight(weight):
for _ in range(N_TRIES):
x_train, x_test, y_train, y_test = \
train_test_split(IRIS.data, IRIS.target,
test_size=0.33, random_state=31)
scikit_model = ScikitRandomForestRegressor(random_state=777)
daal4py_model = DaalRandomForestRegressor(random_state=777)
scikit_predict = scikit_model.fit(
x_train, y_train, sample_weight=weight[0]).predict(x_test)
daal4py_predict = daal4py_model.fit(
x_train, y_train, sample_weight=weight[0]).predict(x_test)
scikit_accuracy = mean_squared_error(scikit_predict, y_test)
daal4py_accuracy = mean_squared_error(daal4py_predict, y_test)
ratio = daal4py_accuracy / scikit_accuracy
assert ratio <= CHECK_RATIO_REGRESSOR, \
('Regression sample weights: sample_weight_type=%s,scikit_accuracy=%f,\
daal4py_accuracy=%f' % (weight[1], scikit_accuracy, daal4py_accuracy))
n_classes=n_classes,
random_state=777)
return train_test_split(x,
y,
random_state=777,
test_size=test_size,
shuffle=shuffle)
ESTIMATORS = {
'KNeighborsClassifier':
KNeighborsClassifier(n_neighbors=10),
'DaalRandomForestClassifier':
DaalRandomForestClassifier(n_estimators=10, random_state=777),
'DaalRandomForestRegressor':
DaalRandomForestRegressor(n_estimators=10, random_state=777),
}
ORDERS = ['C', 'F']
DATA_FORMATS = [pd.DataFrame, np.array]
def check_data_formats_diff(name):
x_train, x_test, y_train, y_test = make_dataset()
alg_results = []
for data_format in DATA_FORMATS:
for order in ORDERS:
x_train_copy = convert_data(x_train.copy(), data_format, order)
x_test_copy = convert_data(x_test.copy(), data_format, order)
y_train_copy = convert_data(y_train.copy(), data_format, order)
y_test_copy = convert_data(y_test.copy(), data_format, order)