def test_scm1d():
df = pd.read_csv('scm1d.csv')
target = df.loc[:, df.columns.str.contains('L')]
df.drop(target.columns, axis=1, inplace=True)
df, target = df.to_numpy(), target.to_numpy()
X_train, X_test, y_train, y_test = train_test_split(df,
target,
test_size=1658.0 /
8145.0,
random_state=42,
shuffle=True)
gb = GradientBoostingRegressor(
l2_regularization=0.07054193143238725,
min_samples_leaf=23,
learning_rate=0.12336530854190006,
max_iter=1999,
n_iter_no_change=None,
)
# scaler = StandardScaler()
# scaler.fit(X_train)
# train_scaled = scaler.transform(X_train)
# test_scaled = scaler.transform(X_test)
# scalery = StandardScaler()
# scalery.fit(y_train)
# y_Train = scalery.transform(y_train)
# y_Test = scalery.transform(y_test)
gb.fit(X_train, y_train)
y_preds = gb.predict_multi(X_test)
r2 = r2_score(y_test, y_preds, multioutput='uniform_average')
print(r2)
def test_atp1d():
df = pd.read_csv('atp1d.csv')
target = df.loc[:, df.columns.str.startswith('LBL')]
df.drop(target.columns, axis=1, inplace=True)
df, target = df.to_numpy(), target.to_numpy()
X_train, X_test, y_train, y_test = train_test_split(df,
target,
test_size=0.5,
random_state=42,
shuffle=True)
gb = GradientBoostingRegressor(l2_regularization=0.003391634274257872,
min_samples_leaf=10,
learning_rate=0.1088115324113492,
max_iter=199,
n_iter_no_change=20)
scaler = StandardScaler()
scaler.fit(X_train)
train_scaled = scaler.transform(X_train)
test_scaled = scaler.transform(X_test)
scalery = StandardScaler()
scalery.fit(y_train)
y_Train = scalery.transform(y_train)
y_Test = scalery.transform(y_test)
gb.fit(train_scaled, y_Train)
y_preds = gb.predict_multi(test_scaled)
r2 = r2_score(y_Test, y_preds, multioutput='uniform_average')
print(r2)
def test_edm():
df = pd.read_csv('edm.csv')
target = df.loc[:, ['DFlow', 'DGap']]
df.drop(target.columns, axis=1, inplace=True)
df, target = df.to_numpy(), target.to_numpy()
X_train, X_test, y_train, y_test = train_test_split(df,
target,
test_size=0.5,
random_state=42,
shuffle=True)
gb = GradientBoostingRegressor(
l2_regularization=0.880826520747869,
min_samples_leaf=12,
learning_rate=0.22445307581959334,
max_iter=279,
n_iter_no_change=23,
)
scaler = StandardScaler()
scaler.fit(X_train)
train_scaled = scaler.transform(X_train)
test_scaled = scaler.transform(X_test)
scalery = StandardScaler()
scalery.fit(y_train)
y_Train = scalery.transform(y_train)
y_Test = scalery.transform(y_test)
gb.fit(train_scaled, y_Train)
y_preds = gb.predict_multi(test_scaled)
r2 = r2_score(y_Test, y_preds, multioutput='uniform_average')
print(r2)
def test_wq():
df = pd.read_csv('water-quality.csv')
target = df.loc[:, df.columns.str.startswith('x')]
df.drop(target.columns, axis=1, inplace=True)
df, target = df.to_numpy(), target.to_numpy()
X_train, X_test, y_train, y_test = train_test_split(df,
target,
test_size=0.5,
random_state=42,
shuffle=True)
gb = GradientBoostingRegressor(l2_regularization=0.07509314619453317,
min_samples_leaf=15,
learning_rate=0.01948991297099692,
max_iter=300,
n_iter_no_change=17)
scaler = StandardScaler()
scaler.fit(X_train)
train_scaled = scaler.transform(X_train)
test_scaled = scaler.transform(X_test)
scalery = StandardScaler()
scalery.fit(y_train)
y_Train = scalery.transform(y_train)
y_Test = scalery.transform(y_test)
gb.fit(train_scaled, y_Train)
y_preds = gb.predict_multi(test_scaled)
r2 = r2_score(y_Test, y_preds, multioutput='uniform_average')
print(r2)
def test_scm20d():
df = pd.read_csv('scm20d.csv')
target = df.loc[:, df.columns.str.contains('L')]
df.drop(target.columns, axis=1, inplace=True)
df, target = df.to_numpy(), target.to_numpy()
c = 1503.0 / 7463.0
X_train, X_test, y_train, y_test = train_test_split(df,
target,
test_size=c,
random_state=42,
shuffle=True)
gb = GradientBoostingRegressor(
l2_regularization=0.8640187696889217,
min_samples_leaf=19,
learning_rate=0.1164232801613771,
max_iter=1998,
n_iter_no_change=None,
)
scaler = StandardScaler()
scaler.fit(X_train)
train_scaled = scaler.transform(X_train)
test_scaled = scaler.transform(X_test)
scalery = StandardScaler()
scalery.fit(y_train)
y_Train = scalery.transform(y_train)
y_Test = scalery.transform(y_test)
gb.fit(train_scaled, y_Train)
y_preds = gb.predict_multi(test_scaled)
r2 = r2_score(y_Test, y_preds, multioutput='uniform_average')
print(r2)
def test_early_stopping_regression(scoring, validation_split, tol):
max_iter = 500
n_iter_no_change = 5
X, y = make_regression(random_state=0)
gb = GradientBoostingRegressor(
verbose=1, # just for coverage
scoring=scoring,
tol=tol,
validation_split=validation_split,
max_iter=max_iter,
n_iter_no_change=n_iter_no_change,
random_state=0)
gb.fit(X, y)
if scoring is not None:
assert n_iter_no_change <= gb.n_iter_ < max_iter
else:
assert gb.n_iter_ == max_iter
def test_same_predictions_regression(seed, min_samples_leaf, n_samples,
max_leaf_nodes):
# Make sure pygbm has the same predictions as LGBM for easy targets.
#
# In particular when the size of the trees are bound and the number of
# samples is large enough, the structure of the prediction trees found by
# LightGBM and PyGBM should be exactly identical.
#
# Notes:
# - Several candidate splits may have equal gains when the number of
# samples in a node is low (and because of float errors). Therefore the
# predictions on the test set might differ if the structure of the tree
# is not exactly the same. To avoid this issue we only compare the
# predictions on the test set when the number of samples is large enough
# and max_leaf_nodes is low enough.
# - To ignore discrepancies caused by small differences the binning
# strategy, data is pre-binned if n_samples > 255.
rng = np.random.RandomState(seed=seed)
n_samples = n_samples
max_iter = 1
max_bins = 256
X, y = make_regression(n_samples=n_samples,
n_features=5,
n_informative=5,
random_state=0)
if n_samples > 255:
# bin data and convert it to float32 so that the estimator doesn't
# treat it as pre-binned
X = BinMapper(max_bins=max_bins).fit_transform(X).astype(np.float32)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=rng)
est_pygbm = GradientBoostingRegressor(max_iter=max_iter,
max_bins=max_bins,
learning_rate=1,
n_iter_no_change=None,
min_samples_leaf=min_samples_leaf,
max_leaf_nodes=max_leaf_nodes)
est_lightgbm = get_lightgbm_estimator(est_pygbm)
est_lightgbm.fit(X_train, y_train)
est_pygbm.fit(X_train, y_train)
# We need X to be treated an numerical data, not pre-binned data.
X_train, X_test = X_train.astype(np.float32), X_test.astype(np.float32)
pred_lgbm = est_lightgbm.predict(X_train)
pred_pygbm = est_pygbm.predict(X_train)
# less than 1% of the predictions are different up to the 3rd decimal
assert np.mean(abs(pred_lgbm - pred_pygbm) > 1e-3) < .011
if max_leaf_nodes < 10 and n_samples >= 1000:
pred_lgbm = est_lightgbm.predict(X_test)
pred_pygbm = est_pygbm.predict(X_test)
# less than 1% of the predictions are different up to the 4th decimal
assert np.mean(abs(pred_lgbm - pred_pygbm) > 1e-4) < .01
def test_pre_binned_data():
# Make sure that:
# - training on numerical data and predicting on numerical data is the
# same as training on binned data and predicting on binned data
# - training on numerical data and predicting on numerical data is the
# same as training on numerical data and predicting on binned data
# - training on binned data and predicting on numerical data is not
# possible.
X, y = make_regression(random_state=0)
gbdt = GradientBoostingRegressor(scoring=None, random_state=0)
mapper = BinMapper(random_state=0)
X_binned = mapper.fit_transform(X)
fit_num_pred_num = gbdt.fit(X, y).predict(X)
fit_binned_pred_binned = gbdt.fit(X_binned, y).predict(X_binned)
fit_num_pred_binned = gbdt.fit(X, y).predict(X_binned)
assert_allclose(fit_num_pred_num, fit_binned_pred_binned)
assert_allclose(fit_num_pred_num, fit_num_pred_binned)
assert_raises_regex(ValueError,
'This estimator was fitted with pre-binned data ',
gbdt.fit(X_binned, y).predict, X)
"""This example illustrates the use of scikit-learn's GridSearchCV.
The grid search is used to determine the best learning rate."""
import numpy as np
from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import KFold
from pygbm import GradientBoostingRegressor
rng = np.random.RandomState(0)
n_samples = int(1e6)
X, y = make_regression(n_samples, random_state=rng)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=rng)
clf = GradientBoostingRegressor(max_iter=10,
scoring=None,
verbose=1,
random_state=rng)
param_grid = {'learning_rate': [1, .1, .01, .001]}
cv = KFold(n_splits=3, random_state=rng)
gs = GridSearchCV(clf, param_grid=param_grid, cv=cv)
gs.fit(X_train, y_train)
print(f'Best param: {gs.best_params_}')
print(f'R2 coefficient: {gs.score(X_test, y_test)}')
"""This example illustrates the use of scikit-learn's GridSearchCV.
The grid search is used to determine the best learning rate."""
import numpy as np
from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import KFold
from pygbm import GradientBoostingRegressor
rng = np.random.RandomState(0)
n_samples = int(1e6)
X, y = make_regression(n_samples, random_state=rng)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=rng)
clf = GradientBoostingRegressor(max_iter=10,
n_iter_no_change=None,
verbose=1,
random_state=rng)
param_grid = {'learning_rate': [1, .1, .01, .001]}
cv = KFold(n_splits=3, random_state=rng)
gs = GridSearchCV(clf, param_grid=param_grid, cv=cv)
gs.fit(X_train, y_train)
print(f'Best param: {gs.best_params_}')
print(f'R2 coefficient: {gs.score(X_test, y_test)}')
