def test_save_pfs():
# Regression
xgboost_regressor_cls = get_xgboost_learner('xgboost.XGBRegressor')
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Model building
path = 'Abalone'
results = features_df.kxy.fit(target_column, xgboost_regressor_cls, \
problem_type='regression', feature_selection_method='pfs', \
path=path)
loaded_predictor = PFSPredictor().load(path, xgboost_regressor_cls)
feature_directions = loaded_predictor.feature_directions
assert feature_directions.shape[1] == features_df.shape[1] - 1
predictions = loaded_predictor.predict(features_df)
assert len(predictions.columns) == 1
assert target_column in predictions.columns
assert set(features_df.index).difference(set(predictions.index)) == set()
assert set(predictions.index).difference(set(features_df.index)) == set()
def test_lean_boosted_xgboost_regressor():
for clz in ['xgboost.XGBRegressor']:
# Regression
xgboost_regressor_cls = get_xgboost_learner(clz)
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Model building
results = features_df.kxy.fit(target_column, xgboost_regressor_cls, \
problem_type='regression', additive_learning=True, return_scores=True, \
n_down_perf_before_stop=1)
model = results['predictor'].models[0]
feature_columns = results['Selected Variables']
x = features_df[feature_columns].values
predictions = model.predict(x)
path = '../kxy/misc/cache/%s-%s.sav' % (dataset.name, clz)
model.save(path)
loaded_model = xgboost_regressor_cls(path=path)
loaded_predictions = loaded_model.predict(x)
assert np.allclose(predictions, loaded_predictions)
def test_rfe_predictor_xgboost():
for clz in ['xgboost.XGBRegressor']:
# Regression
xgboost_regressor_cls = get_xgboost_learner(clz)
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Model building
results = features_df.kxy.fit(target_column, xgboost_regressor_cls, \
problem_type='regression', feature_selection_method='rfe', rfe_n_features=5)
predictor = results['predictor']
predictions = predictor.predict(features_df)
path = '../kxy/misc/cache/%s-%s.sav' % (dataset.name, clz)
predictor.save(path)
loaded_predictor = RFEPredictor.load(path, xgboost_regressor_cls)
loaded_predictions = loaded_predictor.predict(features_df)
assert np.allclose(predictions, loaded_predictions)
def test_xgboost_regression():
regressor_cls = get_xgboost_learner('xgboost.XGBRegressor', random_state=0)
fs = RFE(regressor_cls)
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Feature selection
x_columns = [_ for _ in features_df.columns if _ != target_column]
x_df = features_df[x_columns]
y_df = features_df[[target_column]]
n_vars = max(x_df.shape[1] - 5, 1)
m = fs.fit(x_df, y_df, n_vars)
# Assertions
assert len(fs.selected_variables) == n_vars
assert fs.selected_variables == ['Shell weight', 'Sex_I', 'Shucked weight.ABS(* - Q25(*))', 'Shucked weight', 'Shucked weight.ABS(* - MEDIAN(*))', \
'Shucked weight.ABS(* - MEAN(*))', 'Diameter.ABS(* - Q75(*))', 'Height.ABS(* - Q75(*))', 'Diameter.ABS(* - MEAN(*))', 'Diameter.ABS(* - Q25(*))', \
'Whole weight.ABS(* - MEDIAN(*))', 'Viscera weight.ABS(* - Q75(*))', 'Sex_M', 'Height.ABS(* - MEAN(*))', 'Shucked weight.ABS(* - Q75(*))', \
'Viscera weight.ABS(* - MEAN(*))', 'Height.ABS(* - Q25(*))', 'Whole weight.ABS(* - MEAN(*))', 'Shell weight.ABS(* - Q25(*))', 'Whole weight.ABS(* - Q25(*))', \
'Length.ABS(* - MEAN(*))', 'Length.ABS(* - Q75(*))', 'Whole weight.ABS(* - Q75(*))', 'Diameter.ABS(* - MEDIAN(*))', 'Shell weight.ABS(* - Q75(*))', \
'Shell weight.ABS(* - MEAN(*))', 'Shell weight.ABS(* - MEDIAN(*))', 'Length.ABS(* - MEDIAN(*))', 'Sex_F', 'Viscera weight', 'Whole weight', \
'Length.ABS(* - Q25(*))', 'Viscera weight.ABS(* - MEDIAN(*))']
def test_xgboost_classifier():
# Binary classification
classifier_cls = get_xgboost_learner('xgboost.XGBClassifier',
use_label_encoder=False,
eval_metric='logloss',
learning_rate=0.1,
max_depth=10)
fs = RFE(classifier_cls)
dataset = BankNote()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Feature selection
x_columns = [_ for _ in features_df.columns if _ != target_column]
x_df = features_df[x_columns]
y_df = features_df[[target_column]]
n_vars = max(x_df.shape[1] - 5, 1)
m = fs.fit(x_df, y_df, n_vars)
# Assertions
assert len(fs.selected_variables) == n_vars
assert fs.selected_variables == ['Variance', 'Skewness', 'Kurtosis', 'Entropy', 'Skewness.ABS(* - MEDIAN(*))', \
'Variance.ABS(* - MEAN(*))', 'Skewness.ABS(* - MEAN(*))', 'Kurtosis.ABS(* - MEDIAN(*))', 'Kurtosis.ABS(* - Q25(*))', \
'Entropy.ABS(* - MEDIAN(*))', 'Skewness.ABS(* - Q25(*))', 'Entropy.ABS(* - MEAN(*))', 'Variance.ABS(* - Q25(*))', \
'Kurtosis.ABS(* - MEAN(*))', 'Kurtosis.ABS(* - Q75(*))']
def test_xgboost_regression():
regressor_cls = get_xgboost_learner('xgboost.XGBRegressor', random_state=0)
fs = Boruta(regressor_cls)
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Feature selection
x_columns = [_ for _ in features_df.columns if _ != target_column]
x_df = features_df[x_columns]
y_df = features_df[[target_column]]
m = fs.fit(x_df, y_df)
# Assertions
assert len(fs.selected_variables) == 5
assert fs.selected_variables == [
'Shucked weight', 'Shell weight', 'Sex_I',
'Shucked weight.ABS(* - Q25(*))', 'Whole weight'
]
def test_lean_boosted_xgboost_classifier():
# Binary classification
xgboost_classifier_cls = get_xgboost_learner('xgboost.XGBClassifier',
use_label_encoder=False,
eval_metric='logloss',
learning_rate=0.1,
max_depth=10)
dataset = BankNote()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Model building
results = features_df.kxy.fit(target_column, xgboost_classifier_cls, \
problem_type='classification', additive_learning=True, return_scores=True, \
n_down_perf_before_stop=1)
assert results['Testing Accuracy'] == '0.974'
assert results['Selected Variables'] == [
'Variance', 'Skewness.ABS(* - Q25(*))', 'Kurtosis'
]
def test_boruta():
# Regression
sklearn_regressor_cls = get_xgboost_learner('xgboost.XGBRegressor')
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None, max_lag=None, entity_name='*', exclude=[target_column])
# Model building
results = features_df.kxy.fit(target_column, sklearn_regressor_cls, \
problem_type='regression', feature_selection_method='boruta', boruta_n_evaluations=100)
assert results['Selected Variables'] == ['Shucked weight', 'Shell weight', 'Sex_I', \
'Shucked weight.ABS(* - Q25(*))', 'Whole weight']
def test_single_learner():
# Regression
xgboost_regressor_cls = get_xgboost_learner('xgboost.XGBRegressor')
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Model building
results = features_df.kxy.fit(target_column, xgboost_regressor_cls, \
problem_type='regression', start_n_features=2, min_n_features=2, max_n_features=2, \
additive_learning=True, return_scores=True, n_down_perf_before_stop=1)
assert results['Testing R-Squared'] == '0.493'
assert results['Selected Variables'] == ['Shell weight', 'Shucked weight']
assert len(features_df.kxy.predictor.models) == 1
def test_lean_boosted_xgboost_regressor():
# Regression
xgboost_regressor_cls = get_xgboost_learner('xgboost.XGBRegressor')
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Model building
results = features_df.kxy.fit(target_column, xgboost_regressor_cls, \
problem_type='regression', additive_learning=True, return_scores=True, \
n_down_perf_before_stop=1)
assert results['Testing R-Squared'] == '0.496'
assert results['Selected Variables'] == ['Shell weight', 'Shucked weight', 'Whole weight', \
'Shell weight.ABS(* - Q25(*))', 'Viscera weight.ABS(* - MEDIAN(*))', 'Viscera weight.ABS(* - MEAN(*))', \
'Height', 'Length', 'Diameter', 'Sex_I', 'Shucked weight.ABS(* - MEDIAN(*))', 'Diameter.ABS(* - MEDIAN(*))', \
'Viscera weight.ABS(* - Q75(*))', 'Viscera weight.ABS(* - Q25(*))', 'Diameter.ABS(* - Q25(*))', 'Sex_M', 'Sex_F']
def test_pfs_feature_selection():
# Regression
xgboost_regressor_cls = get_xgboost_learner('xgboost.XGBRegressor')
dataset = Abalone()
target_column = dataset.y_column
df = dataset.df
# Features generation
features_df = df.kxy.generate_features(entity=None,
max_lag=None,
entity_name='*',
exclude=[target_column])
# Model building
results = features_df.kxy.fit(target_column, xgboost_regressor_cls, \
problem_type='regression', feature_selection_method='pfs')
assert results['Feature Directions'].shape[1] == features_df.shape[1] - 1
predictor = results['predictor']
predictions = predictor.predict(features_df)
assert len(predictions.columns) == 1
assert target_column in predictions.columns
assert set(features_df.index).difference(set(predictions.index)) == set()
assert set(predictions.index).difference(set(features_df.index)) == set()
def xgboost_classification_benchmark():
# LeanML vs Boruta vs RFE
xgboost_classifier_cls = get_xgboost_learner('xgboost.XGBClassifier')
classification_benchmark(xgboost_classifier_cls, 'xgboost')
def xgboost_regression_benchmark():
# LeanML vs Boruta vs RFE
xgboost_regressor_cls = get_xgboost_learner('xgboost.XGBRegressor')
regression_benchmark(xgboost_regressor_cls, 'xgboost')