def test_transform(self):
before = pd.DataFrame({
'a': np.random.normal(100, 10, 1000),
'b': np.random.rand(1000),
'c': [0] * 1000,
'd': np.random.randint(-5, 5, 1000),
'e': np.random.rand(1000),
})
# manually add outliers
before['e'][0] = 5
before['e'][1] = -5
before['d'][2] = 100
before['d'][3] = -100
before['b'][4] = 5
before['b'][5] = -5
before['a'][4] = 0
before['a'][5] = 1000
# save means and deviations
means = before.mean()
stds = before.std()
mins = means - 3 * stds
maxes = means + 3 * stds
# fit and transform
clipper = OutliersClipper(['a', 'b', 'c', 'd'])
clipper.fit(before)
after = clipper.transform(before.copy())
# test output
pd.testing.assert_index_equal(before.index, after.index)
pd.testing.assert_series_equal(before['c'], after['c'])
pd.testing.assert_series_equal(before['e'], after['e'])
self.assertEqual(after['d'][2], maxes['d'])
self.assertEqual(after['d'][3], mins['d'])
self.assertFalse((after['d'] > maxes['d']).any())
self.assertFalse((after['d'] < mins['d']).any())
self.assertFalse((after['a'] > maxes['a']).any())
self.assertFalse((after['a'] < mins['a']).any())
self.assertFalse((after['b'] > maxes['b']).any())
self.assertFalse((after['b'] < mins['b']).any())
def get_test_config_houses():
data, labels, continuous, discrete, dummy, categorical, target, missing = get_houses(test=False)
test_data, test_labels = get_houses(test=True)[0:2]
train = data.drop(target, axis=1)
test = test_data.drop(target, axis=1)
scorer = rmse
binner = CustomBinaryBinner({ col: {'values': [train[col].max()]} for col in continuous + discrete })
one_hot = CustomOneHotEncoder(columns=categorical)
model = Pipeline([
('onehot', one_hot),
('clipper', None),
('binner', None),
('binner2', None),
('simple_imputer', None),
('zero_filler', ZeroFiller()),
('main_imputer', None),
('dropper', FeatureDropper(drop=[])),
('poly', None),
('combinations', None),
('boxcox', None),
('scaler', None),
('reduce_dim', None),
('predictor', None)
])
params = {
'LinearRegression_best': {
'params': {'binner2': binner,
'boxcox': BoxCoxTransformer(lambdas_per_column={'BsmtFinSF1': 0, 'LowQualFinSF': 0, 'GrLivArea': 0, 'WoodDeckSF': 0}),
'clipper': None,
'combinations': FeatureProduct(columns=['LotFrontage', 'BsmtFinSF1', 'MasVnrArea', '1stFlrSF', 'GarageArea', 'TotalBsmtSF', 'GrLivArea']),
'dropper__drop': ['LotFrontage_nan', 'MasVnrArea_nan', 'GarageYrBlt_nan'],
'main_imputer': ModelBasedFullImputer(columns=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'],
model=LinearRegression(copy_X=True, fit_intercept=True, n_jobs=7, normalize=False)),
'poly': PolynomialsAdder(powers_per_column={'LotFrontage': [3], 'BsmtFinSF1': [3], 'MasVnrArea': [3], '1stFlrSF': [3], 'GarageArea': [3], 'TotalBsmtSF': [3], 'GrLivArea': [3]}),
'predictor': LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False),
'reduce_dim': PCA(copy=True, iterated_power='auto', n_components=80, random_state=None,
svd_solver='auto', tol=0.0, whiten=False),
'scaler': RobustScaler(copy=True, quantile_range=(25.0, 75.0), with_centering=True,
with_scaling=True),
'simple_imputer': FillNaTransformer(from_dict={}, mean=[], median=[],
nan_flag=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'],
zero=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'])},
'score': 23615.61547841841,
'std': 1575.0296698711768},
# same as baseline
'XGBRegressor_best': {'params': {'binner2': None,
'boxcox': None,
'clipper': None,
'combinations': None,
'dropper__drop': ['LotFrontage_nan', 'MasVnrArea_nan', 'GarageYrBlt_nan'],
'main_imputer': None,
'poly': None,
'predictor': XGBRegressor(
colsample_bytree=1, learning_rate=0.07,
max_depth=3, n_estimators=1000, n_jobs=7),
'reduce_dim': None,
'scaler': None,
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], median=[],
nan_flag=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], zero=[])},
'score': 25011.258367120317,
'std': 1390.9293424644447},
'DecisionTreeRegressor_base': {'params': {'predictor': DecisionTreeRegressor(),
'scaler': None,
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], median=[],
nan_flag=[], zero=[])},
'score': 41942.13985251157,
'std': 4160.532009551353},
'KNeighborsRegressor_base': {'params': {'predictor': KNeighborsRegressor(n_neighbors=7, n_jobs=7),
'scaler': None,
'simple_imputer': FillNaTransformer(from_dict={}, mean=[], median=[], nan_flag=[],
zero=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'])},
'score': 49074.08623384354,
'std': 4863.286944918721},
'LinearRegression_base': {'params': {'predictor': LinearRegression(n_jobs=7),
'scaler': None,
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], median=[],
nan_flag=[], zero=[])},
'score': 58196.855144405956,
'std': 28243.597268210826},
'XGBRegressor_base': {'params': {
'predictor': XGBRegressor(max_depth=3, n_jobs=7),
'scaler': None,
'simple_imputer': FillNaTransformer(from_dict={}, mean=[], median=[], nan_flag=[],
zero=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'])},
'score': 26368.673265668913,
'std': 1310.375215389942},
'XGBRegressor_tuned_base': {'params': {'predictor': XGBRegressor(
colsample_bytree=1,learning_rate=0.07,
max_depth=3, n_estimators=1000, n_jobs=7),
'scaler': None,
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], median=[],
nan_flag=[], zero=[])},
'score': 24894.281304255797,
'std': 1177.011047285853},
'DecisionTreeRegressor_best': {'params': {'binner': None,
'binner2': binner,
'clipper': OutliersClipper(columns=['LotFrontage', 'LotArea', 'MasVnrArea', 'BsmtFinSF1', 'BsmtFinSF2', 'BsmtUnfSF', 'TotalBsmtSF', '1stFlrSF', '2ndFlrSF', 'LowQualFinSF', 'GrLivArea', 'GarageArea', 'WoodDeckSF', 'OpenPorchSF', 'EnclosedPorch', '3SsnPorch', 'ScreenPorch', 'PoolArea', 'MiscVal']),
'combinations': FeatureProduct(columns=['LotFrontage', 'BsmtFinSF1', 'MasVnrArea', '1stFlrSF', 'GarageArea', 'TotalBsmtSF', 'GrLivArea']),
'dropper__drop': ['LotFrontage_nan', 'MasVnrArea_nan', 'GarageYrBlt_nan'],
'main_imputer': HotDeckFullImputer(col_k_pairs=[('LotFrontage', None), ('MasVnrArea', None), ('GarageYrBlt', None)], default_k=5),
'poly': PolynomialsAdder(powers_per_column={'LotFrontage': [2], 'LotArea': [2], 'MasVnrArea': [2], 'BsmtFinSF1': [2], 'BsmtFinSF2': [2], 'BsmtUnfSF': [2], 'TotalBsmtSF': [2], '1stFlrSF': [2], '2ndFlrSF': [2], 'LowQualFinSF': [2], 'GrLivArea': [2], 'GarageArea': [2], 'WoodDeckSF': [2], 'OpenPorchSF': [2], 'EnclosedPorch': [2], '3SsnPorch': [2], 'ScreenPorch': [2], 'PoolArea': [2], 'MiscVal': [2]}),
'predictor': DecisionTreeRegressor(),
'reduce_dim': SelectFromModel(estimator=RandomForestRegressor(max_depth=8)),
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], median=[],
nan_flag=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], zero=[])},
'score': 37866.96889728187,
'std': 5359.25597193946},
'Lasso_best': {'params': {
'binner': None,
'binner2': binner,
'clipper': None,
'combinations': FeatureProduct(
columns=['LotFrontage', 'BsmtFinSF1', 'MasVnrArea', '1stFlrSF', 'GarageArea', 'TotalBsmtSF', 'GrLivArea']),
'dropper__drop': ['LotFrontage_nan', 'MasVnrArea_nan', 'GarageYrBlt_nan'],
'main_imputer': HotDeckFullImputer(col_k_pairs=[('LotFrontage', None), ('MasVnrArea', None), ('GarageYrBlt', None)],
default_k=7),
'poly': None,
'predictor': Lasso(alpha=1.0, copy_X=True, fit_intercept=True, max_iter=1000,
normalize=False, positive=False, precompute=False, random_state=None,
selection='cyclic', tol=0.0001, warm_start=False),
'reduce_dim': PCA(copy=True, iterated_power='auto', n_components=80, random_state=None,
svd_solver='auto', tol=0.0, whiten=False),
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], median=[],
nan_flag=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], zero=[])},
'score': 24138.931080813963,
'std': 639.0998169991468} ,
'Lasso_base': {'params': {
'predictor': Lasso(alpha=1.0, copy_X=True, fit_intercept=True, max_iter=1000,
normalize=False, positive=False, precompute=False, random_state=None,
selection='cyclic', tol=0.0001, warm_start=False),
'scaler': RobustScaler(),
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'])},
'score': 0,
'std': 0}
}
return data, test, test_labels, scorer, model, params, target
def get_test_config_boston(missing=True):
data, labels, continuous, discrete, dummy, categorical, target, missing = get_boston(
test=False, missing=missing)
test_data, test_labels = get_boston(test=True, missing=missing)[0:2]
test = test_data.drop(target, axis=1)
scorer = rmse
model = Pipeline([('clipper', None), ('binner', None), ('binner2', None),
('simple_imputer', None), ('zero_filler', ZeroFiller()),
('main_imputer', None),
('dropper', FeatureDropper(drop=[])), ('poly', None),
('combinations', None), ('boxcox', None),
('scaler', None), ('reduce_dim', None),
('predictor', None)])
params = {
# 0
'XGBRegressor_best': {
'params': {
'binner2':
None,
'boxcox':
BoxCoxTransformer(lambdas_per_column={
'age': 2,
'tax': 0,
'lstat': 0
}),
'clipper':
OutliersClipper(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'lstat', 'dis'
]),
'combinations':
None,
'dropper__drop': [],
'main_imputer':
ModelBasedFullImputer(
columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'dis'
],
model=DecisionTreeRegressor(max_depth=None)),
'poly':
PolynomialsAdder(
powers_per_column={
'crim': [2, 3],
'zn': [2, 3],
'nox': [2, 3],
'indus': [2, 3],
'rm': [2, 3],
'age': [2, 3],
'tax': [2, 3],
'ptratio': [2, 3],
'b': [2, 3],
'lstat': [2, 3],
'dis': [2, 3]
}),
'predictor':
XGBRegressor(learning_rate=0.05, max_depth=6,
n_estimators=500),
'reduce_dim':
None,
'scaler':
None,
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[],
median=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
nan_flag=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
zero=[])
},
'score': 3.7821358047127682,
'std': 0.4967512627490983
},
# 0
'Lasso_best': {
'params': {
'binner2':
None,
'boxcox':
BoxCoxTransformer(lambdas_per_column={
'age': 2,
'tax': 0,
'lstat': 0
}),
'clipper':
OutliersClipper(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'lstat', 'dis'
]),
'combinations':
FeatureProduct(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'lstat', 'dis'
]),
'dropper__drop': [
'crim_nan', 'zn_nan', 'nox_nan', 'indus_nan', 'rm_nan',
'age_nan', 'tax_nan', 'ptratio_nan', 'b_nan', 'dis_nan'
],
'main_imputer':
ModelBasedFullImputer(
columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'dis'
],
model=DecisionTreeRegressor(max_depth=None)),
'poly':
None,
'predictor':
Lasso(alpha=0.01),
'reduce_dim':
None,
'scaler':
RobustScaler(),
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[],
median=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
nan_flag=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
zero=[])
},
'score': 3.993797473454735,
'std': 0.5808956921355953
},
# 0
'LinearRegression_best': {
'params': {
'binner2':
None,
'boxcox':
BoxCoxTransformer(lambdas_per_column={
'age': 2,
'tax': 0,
'lstat': 0
}),
'clipper':
OutliersClipper(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'lstat', 'dis'
]),
'combinations':
None,
'dropper__drop': [
'crim_nan', 'zn_nan', 'nox_nan', 'indus_nan', 'rm_nan',
'age_nan', 'tax_nan', 'ptratio_nan', 'b_nan', 'dis_nan'
],
'main_imputer':
ModelBasedFullImputer(
columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'dis'
],
model=DecisionTreeRegressor(max_depth=8)),
'poly':
PolynomialsAdder(
powers_per_column={
'crim': [2, 3],
'zn': [2, 3],
'nox': [2, 3],
'indus': [2, 3],
'rm': [2, 3],
'age': [2, 3],
'tax': [2, 3],
'ptratio': [2, 3],
'b': [2, 3],
'lstat': [2, 3],
'dis': [2, 3]
}),
'predictor':
LinearRegression(),
'reduce_dim':
None,
'scaler':
None,
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[],
median=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
nan_flag=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
zero=[])
},
'score': 4.514645815970899,
'std': 0.7631593234069367
},
'DecisionTreeRegressor_base': {
'params': {
'predictor':
DecisionTreeRegressor(criterion='mse',
max_depth=4,
max_features=None,
max_leaf_nodes=None,
min_impurity_decrease=0.0,
min_impurity_split=None,
min_samples_leaf=1,
min_samples_split=2,
min_weight_fraction_leaf=0.0,
presort=False,
random_state=None,
splitter='best'),
'scaler':
None,
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[],
median=[],
nan_flag=[],
zero=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
])
},
'score': 5.5088106991425985,
'std': 0.293662905734789
},
'KNeighborsRegressor_base': {
'params': {
'predictor':
KNeighborsRegressor(algorithm='auto',
leaf_size=30,
metric='minkowski',
metric_params=None,
n_jobs=1,
n_neighbors=7,
p=2,
weights='uniform'),
'scaler':
RobustScaler(copy=True,
quantile_range=(25.0, 75.0),
with_centering=True,
with_scaling=True),
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
median=[],
nan_flag=[],
zero=[])
},
'score': 5.859771905373064,
'std': 0.90721907618626
},
'LinearRegression_base': {
'params': {
'predictor':
LinearRegression(copy_X=True,
fit_intercept=True,
n_jobs=1,
normalize=False),
'scaler':
RobustScaler(copy=True,
quantile_range=(25.0, 75.0),
with_centering=True,
with_scaling=True),
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[],
median=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
nan_flag=[],
zero=[])
},
'score': 5.494688479501426,
'std': 0.5377531716144219
},
'Lasso_base': {
'params': {
'predictor':
Lasso(alpha=0.01),
'scaler':
RobustScaler(copy=True,
quantile_range=(25.0, 75.0),
with_centering=True,
with_scaling=True),
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[],
median=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
nan_flag=[],
zero=[])
},
'score': 0,
'std': 0
},
# 'XGBRegressor_base': {'params': {
# 'predictor': XGBRegressor(max_depth=4),
# 'scaler': None,
# 'simple_imputer': FillNaTransformer(from_dict={},
# mean=['crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax', 'ptratio', 'b', 'dis'],
# median=[], nan_flag=[], zero=[])},
# 'score': 4.088217989236429,
# 'std': 0.5303490714753816,
# },
'XGBRegressor_tuned_base': {
'params': {
'predictor':
XGBRegressor(learning_rate=0.05, max_depth=6,
n_estimators=500),
'scaler':
None,
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
median=[],
nan_flag=[],
zero=[])
},
'score': 3.942697483564859,
'std': 0.6029251513214098
},
'DecisionTreeRegressor_best': {
'params': {
'binner2':
None,
'boxcox':
BoxCoxTransformer(lambdas_per_column={
'age': 2,
'tax': 0,
'lstat': 0
}),
'clipper':
OutliersClipper(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'lstat', 'dis'
]),
'combinations':
None,
'dropper__drop': [],
'main_imputer':
ModelBasedFullImputer(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax',
'ptratio', 'b', 'dis'
],
model=DecisionTreeRegressor(
criterion='mse',
max_depth=None,
max_features=None,
max_leaf_nodes=None,
min_impurity_decrease=0.0,
min_impurity_split=None,
min_samples_leaf=1,
min_samples_split=2,
min_weight_fraction_leaf=0.0,
presort=False,
random_state=None,
splitter='best')),
'poly':
None,
'predictor':
DecisionTreeRegressor(criterion='mse',
max_depth=8,
max_features=None,
max_leaf_nodes=None,
min_impurity_decrease=0.0,
min_impurity_split=None,
min_samples_leaf=1,
min_samples_split=2,
min_weight_fraction_leaf=0.0,
presort=False,
random_state=None,
splitter='best'),
'reduce_dim':
None,
'scaler':
None,
'simple_imputer':
FillNaTransformer(from_dict={},
mean=[],
median=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
nan_flag=[
'crim', 'zn', 'nox', 'indus', 'rm',
'age', 'tax', 'ptratio', 'b', 'dis'
],
zero=[])
},
'score': 4.840076015850126,
'std': 0.718619669114889
},
}
return data, test, test_labels, scorer, model, params, target
def get_test_config_heart(missing=True):
data, labels, continuous, discrete, dummy, categorical, target, missing = get_heart(
test=False, missing=missing)
test_data, test_labels = get_heart(test=True, missing=missing)[0:2]
test = test_data.drop(target, axis=1)
scorer = error_rate
one_hot = CustomOneHotEncoder(columns=categorical)
model = Pipeline([
('onehot', one_hot),
('clipper', None),
('binner', None),
('binner2', None),
('simple_imputer', None),
('zero_filler', ZeroFiller()), # just in case there are any left
('main_imputer', None),
('dropper', FeatureDropper(drop=[])),
('poly', None),
('combinations', None),
('boxcox', None),
('scaler', None),
('reduce_dim', None),
('predictor', None)
])
params = {
'DecisionTreeClassifier_base': {
'params': {
'predictor':
DecisionTreeClassifier(max_depth=None),
'scaler':
None,
'simple_imputer':
FillNaTransformer(
from_dict={},
mean=['trestbps', 'chol', 'thalach', 'oldpeak'],
median=[],
nan_flag=[],
zero=[])
},
'score': 0.1974390243902439,
'std': 0.0756691348271984
},
'KNeighborsClassifier_base': {
'params': {
'predictor':
KNeighborsClassifier(n_neighbors=7
# ,n_jobs=7
),
'scaler':
RobustScaler(),
'simple_imputer':
FillNaTransformer(
from_dict={},
mean=[],
median=['trestbps', 'chol', 'thalach', 'oldpeak'],
nan_flag=[],
zero=[])
},
'score': 0.1924390243902439,
'std': 0.04087385740197896
},
'LogisticRegression_base': {
'params': {
'predictor':
LogisticRegression(
# n_jobs=7,
),
'scaler':
None,
'simple_imputer':
FillNaTransformer(
from_dict={},
mean=['trestbps', 'chol', 'thalach', 'oldpeak'],
median=[],
nan_flag=[],
zero=[])
},
'score': 0.1825609756097561,
'std': 0.04141604180434009
},
# 'XGBClassifier_base': {'params': {'predictor': XGBClassifier(
# base_score=0.5,
# # n_jobs=7,
#
# colsample_bytree=0.8, learning_rate=0.07,
# max_depth=7, n_estimators=200,),
# 'scaler': None,
# 'simple_imputer': FillNaTransformer(from_dict={}, mean=[], median=[], nan_flag=[],
# zero=['trestbps', 'chol', 'thalach', 'oldpeak'])},
# 'score': 0.16743902439024388,
# 'std': 0.04176646782554455},
'DecisionTreeClassifier_best': {
'params': {
'binner2':
CustomBinner(
configuration={
'chol': {
'bins': 3
},
'thalach': {
'bins': 3
},
'oldpeak': {
'bins': 3
},
'trestbps': {
'bins': 3
},
'age': {
'bins': 3
},
'slope': {
'bins': 3
},
'ca': {
'bins': 3
}
}),
'boxcox':
None,
'clipper':
OutliersClipper(
columns=['chol', 'thalach', 'oldpeak', 'trestbps']),
'combinations':
FeatureProduct(
columns=['chol', 'thalach', 'oldpeak', 'trestbps']),
'dropper__drop':
['trestbps_nan', 'chol_nan', 'thalach_nan', 'oldpeak_nan'],
'main_imputer':
HotDeckFullImputer(col_k_pairs=[('trestbps', None),
('chol', None),
('thalach', None),
('oldpeak', None)],
default_k=7),
'poly':
None,
'predictor':
DecisionTreeClassifier(max_depth=4),
'reduce_dim':
None,
'scaler':
None,
'simple_imputer':
FillNaTransformer(
from_dict={},
mean=[],
median=[],
nan_flag=['trestbps', 'chol', 'thalach', 'oldpeak'],
zero=['trestbps', 'chol', 'thalach', 'oldpeak'])
},
'score': 0.14780487804878048,
'std': 0.03090350740255695
},
'LogisticRegression_best': {
'params': {
'binner2':
None,
'boxcox':
BoxCoxTransformer(lambdas_per_column={
'chol': 0,
'thalach': 2,
'trestbps': 0
}),
'clipper':
OutliersClipper(
columns=['chol', 'thalach', 'oldpeak', 'trestbps']),
'combinations':
None,
'dropper__drop': [],
'main_imputer':
ModelBasedFullImputer(
columns=['trestbps', 'chol', 'thalach', 'oldpeak'],
model=LinearRegression(
# n_jobs=7
)),
'poly':
PolynomialsAdder(powers_per_column={
'chol': [2],
'thalach': [2],
'oldpeak': [2],
'trestbps': [2]
}),
'predictor':
LogisticRegression(),
'reduce_dim':
PCA(n_components=10),
'scaler':
None,
'simple_imputer':
FillNaTransformer(
from_dict={},
mean=['trestbps', 'chol', 'thalach', 'oldpeak'],
median=[],
nan_flag=['trestbps', 'chol', 'thalach', 'oldpeak'],
zero=[])
},
'score': 0.14280487804878048,
'std': 0.03915450868377355
},
# 'XGBClassifier_best': {'params':
# {
# 'binner2': CustomBinner(configuration={'chol': {'bins': 3}, 'thalach': {'bins': 3}, 'oldpeak': {'bins': 3}, 'trestbps': {'bins': 3}, 'age': {'bins': 3}, 'slope': {'bins': 3}, 'ca': {'bins': 3}},
# drop=False, nan=False),
# 'boxcox': BoxCoxTransformer(lambdas_per_column={'chol': 0, 'thalach': 2, 'trestbps': 0}),
# 'clipper': OutliersClipper(columns=['chol', 'thalach', 'oldpeak', 'trestbps']),
# 'combinations': None,
# 'dropper__drop': ['trestbps_nan', 'chol_nan', 'thalach_nan', 'oldpeak_nan'],
# 'main_imputer': HotDeckFullImputer(col_k_pairs=[('trestbps', None), ('chol', None), ('thalach', None), ('oldpeak', None)],
# default_k=7),
# 'poly': None,
# 'predictor': XGBClassifier(
# # n_jobs=7,
# base_score=0.5,
# colsample_bytree=0.8, learning_rate=0.07,
# max_depth=7, n_estimators=200,),
# 'reduce_dim': SelectFromModel(estimator=LogisticRegression(C=0.999, penalty='l1',
# # n_jobs=7
# )),
# 'scaler': None,
# 'simple_imputer': FillNaTransformer(from_dict={},
# mean=['trestbps', 'chol', 'thalach', 'oldpeak'], median=[],
# nan_flag=['trestbps', 'chol', 'thalach', 'oldpeak'], zero=[])},
# 'score': 0.15243902439024387,
# 'std': 0.04655758333858798}
}
return data, test, test_labels, scorer, model, params, target
'bins': 3
},
'age': {
'bins': 3
},
'slope': {
'bins': 3
},
'ca': {
'bins': 3
}
}),
'boxcox':
None,
'clipper':
OutliersClipper(
columns=['chol', 'thalach', 'oldpeak', 'trestbps']),
'combinations':
FeatureProduct(columns=['chol', 'thalach', 'oldpeak', 'trestbps']),
'dropper__drop':
['trestbps_nan', 'chol_nan', 'thalach_nan', 'oldpeak_nan'],
'main_imputer':
HotDeckFullImputer(col_k_pairs=[('trestbps', None), ('chol', None),
('thalach', None),
('oldpeak', None)],
default_k=7),
'poly':
None,
'predictor':
DecisionTreeClassifier(max_depth=4),
'reduce_dim':
None,
'std': 1310.375215389942},
'XGBRegressor_tuned': {'params': {'predictor': XGBRegressor(
colsample_bytree=1,learning_rate=0.07,
max_depth=3, n_estimators=1000),
'scaler': None,
'simple_imputer': FillNaTransformer(from_dict={},
mean=['LotFrontage', 'MasVnrArea', 'GarageYrBlt'], median=[],
nan_flag=[], zero=[])},
'score': 24894.281304255797,
'std': 1177.011047285853}
}
best2 = {'DecisionTreeRegressor': {'params': {'binner': None,
'binner2': CustomBinaryBinner(configuration={'LotFrontage': {'values': [182.0]}, 'LotArea': {'values': [215245]}, 'MasVnrArea': {'values': [1378.0]}, 'BsmtFinSF1': {'values': [2188]}, 'BsmtFinSF2': {'values': [1120]}, 'BsmtUnfSF': {'values': [2336]}, 'TotalBsmtSF': {'values': [3206]}, '1stFlrSF': {'values': [3228]}, '2ndFlrSF': ... [2010.0]}, 'GarageCars': {'values': [4]}, 'MoSold': {'values': [12]}, 'YrSold': {'values': [2010]}},
drop=False, nan=False),
'clipper': OutliersClipper(columns=['LotFrontage', 'LotArea', 'MasVnrArea', 'BsmtFinSF1', 'BsmtFinSF2', 'BsmtUnfSF', 'TotalBsmtSF', '1stFlrSF', '2ndFlrSF', 'LowQualFinSF', 'GrLivArea', 'GarageArea', 'WoodDeckSF', 'OpenPorchSF', 'EnclosedPorch', '3SsnPorch', 'ScreenPorch', 'PoolArea', 'MiscVal']),
'combinations': FeatureProduct(columns=['LotFrontage', 'BsmtFinSF1', 'MasVnrArea', '1stFlrSF', 'GarageArea', 'TotalBsmtSF', 'GrLivArea']),
'dropper__drop': ['LotFrontage_nan', 'MasVnrArea_nan', 'GarageYrBlt_nan'],
'main_imputer': HotDeckFullImputer(col_k_pairs=[('LotFrontage', None), ('MasVnrArea', None), ('GarageYrBlt', None)],
default_k=5),
'poly': PolynomialsAdder(powers_per_column={'LotFrontage': [2], 'LotArea': [2], 'MasVnrArea': [2], 'BsmtFinSF1': [2], 'BsmtFinSF2': [2], 'BsmtUnfSF': [2], 'TotalBsmtSF': [2], '1stFlrSF': [2], '2ndFlrSF': [2], 'LowQualFinSF': [2], 'GrLivArea': [2], 'GarageArea': [2], 'WoodDeckSF': [2], 'OpenPorchSF': [2], 'EnclosedPorch': [2], '3SsnPorch': [2], 'ScreenPorch': [2], 'PoolArea': [2], 'MiscVal': [2]}),
'predictor': DecisionTreeRegressor(criterion='mse', max_depth=None, max_features=None,
max_leaf_nodes=None, min_impurity_decrease=0.0,
min_impurity_split=None, min_samples_leaf=1,
min_samples_split=2, min_weight_fraction_leaf=0.0,
presort=False, random_state=None, splitter='best'),
'reduce_dim': SelectFromModel(estimator=RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=8,
max_features='auto', max_leaf_nodes=None,
min_impurity_decrease=0.0, min_impurity_split=None,
min_samples_leaf=1, min_samples_split=2,
min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,
best = {
# 0
'XGBRegressor': {
'params': {
'binner2':
None,
'boxcox':
BoxCoxTransformer(lambdas_per_column={
'age': 2,
'tax': 0,
'lstat': 0
}),
'clipper':
OutliersClipper(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax', 'ptratio',
'b', 'lstat', 'dis'
]),
'combinations':
None,
'dropper__drop': [],
'main_imputer':
ModelBasedFullImputer(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax', 'ptratio',
'b', 'dis'
],
model=DecisionTreeRegressor(max_depth=None)),
'poly':
PolynomialsAdder(
powers_per_column={
'crim': [2, 3],
'zn': [2, 3],
'trestbps': 0
}),
'clipper':
None,
'scaler':
None
}
BASE_BOSTON = {
'binner':
CustomBinaryBinner(configuration={}),
'boxcox':
None,
'clipper':
OutliersClipper(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax', 'ptratio', 'b',
'lstat'
]),
'scaler':
StandardScaler(copy=True, with_mean=True, with_std=True)
}
BASE_HOUSES = {
'binner':
CustomBinaryBinner(configuration={}),
'boxcox':
None,
'clipper':
OutliersClipper(columns=[
'crim', 'zn', 'nox', 'indus', 'rm', 'age', 'tax', 'ptratio', 'b',
'lstat'
]),