def test_early_stopping_classification(
data, scoring, validation_fraction, early_stopping, n_iter_no_change, tol
):
max_iter = 50
X, y = data
gb = HistGradientBoostingClassifier(
verbose=1, # just for coverage
min_samples_leaf=5, # easier to overfit fast
scoring=scoring,
tol=tol,
early_stopping=early_stopping,
validation_fraction=validation_fraction,
max_iter=max_iter,
n_iter_no_change=n_iter_no_change,
random_state=0,
)
gb.fit(X, y)
if early_stopping is True:
assert n_iter_no_change <= gb.n_iter_ < max_iter
else:
assert gb.n_iter_ == max_iter
def test_binning_train_validation_are_separated():
# Make sure training and validation data are binned separately.
# See issue 13926
rng = np.random.RandomState(0)
validation_fraction = .2
gb = HistGradientBoostingClassifier(
n_iter_no_change=5,
validation_fraction=validation_fraction,
random_state=rng)
gb.fit(X_classification, y_classification)
mapper_training_data = gb.bin_mapper_
# Note that since the data is small there is no subsampling and the
# random_state doesn't matter
mapper_whole_data = _BinMapper(random_state=0)
mapper_whole_data.fit(X_classification)
n_samples = X_classification.shape[0]
assert np.all(
mapper_training_data.actual_n_bins_ == int((1 - validation_fraction) *
n_samples))
assert np.all(
mapper_training_data.actual_n_bins_ != mapper_whole_data.actual_n_bins_
)
def hist_gradient_boosting_classifier(x_train, y_train, x_test, y_test):
from sklearn.experimental import enable_hist_gradient_boosting
from sklearn.ensemble import HistGradientBoostingClassifier
ensem = HistGradientBoostingClassifier()
ensem.fit(x_train, y_train)
value = ensem.score(x_test, y_test)
return "{0:.2f}".format(value)
def test_invalid_classification_loss():
binary_clf = HistGradientBoostingClassifier(loss="binary_crossentropy")
err_msg = ("loss='binary_crossentropy' is not defined for multiclass "
"classification with n_classes=3, use "
"loss='categorical_crossentropy' instead")
with pytest.raises(ValueError, match=err_msg):
binary_clf.fit(np.zeros(shape=(3, 2)), np.arange(3))
def test_invalid_classification_loss():
binary_clf = HistGradientBoostingClassifier(loss="binary_crossentropy")
err_msg = ("loss='binary_crossentropy' is not defined for multiclass "
"classification with n_classes=3, use "
"loss='categorical_crossentropy' instead")
with pytest.raises(ValueError, match=err_msg):
binary_clf.fit(np.zeros(shape=(3, 2)), np.arange(3))
def test_zero_sample_weights_classification():
# Make sure setting a SW to zero amounts to ignoring the corresponding
# sample
X = [[1, 0],
[1, 0],
[1, 0],
[0, 1]]
y = [0, 0, 1, 0]
# ignore the first 2 training samples by setting their weight to 0
sample_weight = [0, 0, 1, 1]
gb = HistGradientBoostingClassifier(loss='binary_crossentropy',
min_samples_leaf=1)
gb.fit(X, y, sample_weight=sample_weight)
assert_array_equal(gb.predict([[1, 0]]), [1])
X = [[1, 0],
[1, 0],
[1, 0],
[0, 1],
[1, 1]]
y = [0, 0, 1, 0, 2]
# ignore the first 2 training samples by setting their weight to 0
sample_weight = [0, 0, 1, 1, 1]
gb = HistGradientBoostingClassifier(loss='categorical_crossentropy',
min_samples_leaf=1)
gb.fit(X, y, sample_weight=sample_weight)
assert_array_equal(gb.predict([[1, 0]]), [1])
def test_same_predictions_classification(seed, min_samples_leaf, n_samples,
max_leaf_nodes):
# Same as test_same_predictions_regression but for classification
pytest.importorskip("lightgbm")
rng = np.random.RandomState(seed=seed)
max_iter = 1
max_bins = 255
X, y = make_classification(
n_samples=n_samples,
n_classes=2,
n_features=5,
n_informative=5,
n_redundant=0,
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(n_bins=max_bins + 1).fit_transform(X).astype(np.float32)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=rng)
est_sklearn = HistGradientBoostingClassifier(
loss="binary_crossentropy",
max_iter=max_iter,
max_bins=max_bins,
learning_rate=1,
early_stopping=False,
min_samples_leaf=min_samples_leaf,
max_leaf_nodes=max_leaf_nodes,
)
est_lightgbm = get_equivalent_estimator(est_sklearn, lib="lightgbm")
est_lightgbm.fit(X_train, y_train)
est_sklearn.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_lightgbm = est_lightgbm.predict(X_train)
pred_sklearn = est_sklearn.predict(X_train)
assert np.mean(pred_sklearn == pred_lightgbm) > 0.89
acc_lightgbm = accuracy_score(y_train, pred_lightgbm)
acc_sklearn = accuracy_score(y_train, pred_sklearn)
np.testing.assert_almost_equal(acc_lightgbm, acc_sklearn)
if max_leaf_nodes < 10 and n_samples >= 1000:
pred_lightgbm = est_lightgbm.predict(X_test)
pred_sklearn = est_sklearn.predict(X_test)
assert np.mean(pred_sklearn == pred_lightgbm) > 0.89
acc_lightgbm = accuracy_score(y_test, pred_lightgbm)
acc_sklearn = accuracy_score(y_test, pred_sklearn)
np.testing.assert_almost_equal(acc_lightgbm, acc_sklearn, decimal=2)
def test_string_target_early_stopping(scoring):
# Regression tests for #14709 where the targets need to be encoded before
# to compute the score
rng = np.random.RandomState(42)
X = rng.randn(100, 10)
y = np.array(["x"] * 50 + ["y"] * 50, dtype=object)
gbrt = HistGradientBoostingClassifier(n_iter_no_change=10, scoring=scoring)
gbrt.fit(X, y)
def histGradientModel(X_train,Y_train):
# use Hist Gradient Boosting Classifier
from sklearn.experimental import enable_hist_gradient_boosting
from sklearn.ensemble import HistGradientBoostingClassifier
histgrad=HistGradientBoostingClassifier()
histgrad.fit(X_train,y_train)
print('\nHist Gradient Boosting Training Score:',histgrad.score(X_train,Y_train))
return histgrad,histgrad.score(X_train,Y_train)
def common_test_model_hgb_classifier(self, add_nan=False, n_classes=2):
model = HistGradientBoostingClassifier(max_iter=5, max_depth=2)
X, y = make_classification(n_features=10,
n_samples=1000,
n_informative=4,
n_classes=n_classes,
random_state=42)
if add_nan:
rows = numpy.random.randint(0, X.shape[0] - 1, X.shape[0] // 3)
cols = numpy.random.randint(0, X.shape[1] - 1, X.shape[0] // 3)
X[rows, cols] = numpy.nan
X_train, X_test, y_train, _ = train_test_split(X,
y,
test_size=0.5,
random_state=42)
model.fit(X_train, y_train)
model_onnx = convert_sklearn(
model, "unused", [("input", FloatTensorType([None, X.shape[1]]))])
self.assertIsNotNone(model_onnx)
X_test = X_test.astype(numpy.float32)[:5]
dump_data_and_model(X_test,
model,
model_onnx,
basename="SklearnHGBClassifier%s%d" %
("nan" if add_nan else '', n_classes),
verbose=False,
allow_failure="StrictVersion(onnx.__version__)"
" < StrictVersion('1.2') or "
"StrictVersion(onnxruntime.__version__)"
" <= StrictVersion('0.2.1')")
if n_classes == 2:
model_onnx = convert_sklearn(
model,
"unused", [("input", FloatTensorType([None, X.shape[1]]))],
options={model.__class__: {
'raw_scores': True
}})
self.assertIsNotNone(model_onnx)
X_test = X_test.astype(numpy.float32)[:5]
# There is a bug in onnxruntime <= 1.1.0.
# Raw scores are always positive.
dump_data_and_model(
X_test,
model,
model_onnx,
basename="SklearnHGBClassifierRaw%s%d" %
("nan" if add_nan else '', n_classes),
verbose=False,
allow_failure="StrictVersion(onnx.__version__)"
" < StrictVersion('1.2') or "
"StrictVersion(onnxruntime.__version__)"
" < StrictVersion('1.2.0')",
methods=['predict', 'decision_function_binary'])
def test_crossentropy_binary_problem():
# categorical_crossentropy should only be used if there are more than two
# classes present. PR #14869
X = [[1], [0]]
y = [0, 1]
gbrt = HistGradientBoostingClassifier(loss='categorical_crossentropy')
with pytest.raises(ValueError,
match="'categorical_crossentropy' is not suitable for"):
gbrt.fit(X, y)
def model_(X, y, rs):
osp = RandomUnderSampler(random_state=rs)
x_train_, y_train_ = osp.fit_sample(X, y)
# # 基础模型
# clf = CatBoostClassifier(loss_function='Logloss',
# logging_level='Silent',
# cat_features=categorical_features_indices)
clf = HistGradientBoostingClassifier(random_state=10)
clf.fit(x_train_, y_train_)
return clf
def k_fold_trainning(rawdata,n_folds=5):
cv = StratifiedKFold(n_splits=n_folds,shuffle=True)
target = np.array(rawdata[0].values)
lure = np.array(rawdata[1].values)
y = np.array(rawdata['label'].values)
tprs = []
aucs = []
mean_fpr = np.linspace(0, 1, 100)
fig, ax = plt.subplots()
for i, (train, test) in enumerate(cv.split(target, lure, y)):
print('----------------Training Fold %d---------------'%(i+1))
X_train = pd.DataFrame({0:target[train],1:lure[train]})
X_test = pd.DataFrame({0:target[test],1:lure[test]})
pmfm = create_pmfm(X_train,y[train])
train_feature = X_train.apply(feature_Encoding, axis = 1, args = (0,1,pmfm)).values
test_feature = X_test.apply(feature_Encoding, axis = 1, args = (0,1,pmfm)).values
clf = HistGradientBoostingClassifier(max_iter=500, learning_rate=0.9, max_depth=6, l2_regularization=100)
train_data = np.matrix([train_feature[i] for i in range(train_feature.shape[0])])
test_data = np.matrix([test_feature[i] for i in range(test_feature.shape[0])])
clf.fit(train_data, y[train])
pred = clf.predict(test_data)
evaluate(y[test], pred)
viz = plot_roc_curve(clf, test_data, y[test],
name='ROC fold {}'.format(i+1),
alpha=0.5, lw=1, ax=ax)
interp_tpr = np.interp(mean_fpr, viz.fpr, viz.tpr)
interp_tpr[0] = 0.0
tprs.append(interp_tpr)
aucs.append(viz.roc_auc)
ax.plot([0, 1], [0, 1], linestyle='--', lw=2, color='r',
label='Chance', alpha=.8)
mean_tpr = np.mean(tprs, axis=0)
mean_tpr[-1] = 1.0
mean_auc = auc(mean_fpr, mean_tpr)
std_auc = np.std(aucs)
ax.plot(mean_fpr, mean_tpr, color='b',
label=r'Mean ROC (AUC = %0.3f $\pm$ %0.3f)' % (mean_auc, std_auc),
lw=2, alpha=.8)
std_tpr = np.std(tprs, axis=0)
tprs_upper = np.minimum(mean_tpr + std_tpr, 1)
tprs_lower = np.maximum(mean_tpr - std_tpr, 0)
ax.fill_between(mean_fpr, tprs_lower, tprs_upper, color='grey', alpha=.2,
label=r'$\pm$ 1 std. dev.')
ax.set(xlim=[-0.05, 1.05], ylim=[-0.05, 1.05],
title="Receiver operating characteristic Curve")
ax.legend(loc="lower right")
plt.savefig('roc.png',dpi=300)
def train_model(rawdata, compute_importance=True):
X,y = create_Input(rawdata)
pmfm = create_pmfm(X,y)
np.save("feature.npy",pmfm)
feature = X.apply(feature_Encoding, axis = 1, args = (0,1,pmfm)).values
clf = HistGradientBoostingClassifier(max_iter=500, learning_rate=0.9, max_depth=6, l2_regularization=100)
data = np.matrix([feature[i] for i in range(feature.shape[0])])
clf.fit(data, y)
joblib.dump(clf, "train_model.pkl")
if compute_importance:
feature_importance(clf, data, y)
def test_zero_division_hessians(data):
# non regression test for issue #14018
# make sure we avoid zero division errors when computing the leaves values.
# If the learning rate is too high, the raw predictions are bad and will
# saturate the softmax (or sigmoid in binary classif). This leads to
# probabilities being exactly 0 or 1, gradients being constant, and
# hessians being zero.
X, y = data
gb = HistGradientBoostingClassifier(learning_rate=100, max_iter=10)
gb.fit(X, y)
def test_early_stopping_on_test_set_with_warm_start():
# Non regression test for #16661 where second fit fails with
# warm_start=True, early_stopping is on, and no validation set
X, y = make_classification(random_state=0)
gb = HistGradientBoostingClassifier(
max_iter=1, scoring='loss', warm_start=True, early_stopping=True,
n_iter_no_change=1, validation_fraction=None)
gb.fit(X, y)
# does not raise on second call
gb.set_params(max_iter=2)
gb.fit(X, y)
def test_same_predictions_classification(seed, min_samples_leaf, n_samples,
max_leaf_nodes):
# Same as test_same_predictions_regression but for classification
rng = np.random.RandomState(seed=seed)
n_samples = n_samples
max_iter = 1
max_bins = 256
X, y = make_classification(n_samples=n_samples, n_classes=2, n_features=5,
n_informative=5, n_redundant=0, 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_sklearn = HistGradientBoostingClassifier(
loss='binary_crossentropy',
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_equivalent_estimator(est_sklearn, lib='lightgbm')
est_lightgbm.fit(X_train, y_train)
est_sklearn.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_lightgbm = est_lightgbm.predict(X_train)
pred_sklearn = est_sklearn.predict(X_train)
assert np.mean(pred_sklearn == pred_lightgbm) > .89
acc_lightgbm = accuracy_score(y_train, pred_lightgbm)
acc_sklearn = accuracy_score(y_train, pred_sklearn)
np.testing.assert_almost_equal(acc_lightgbm, acc_sklearn)
if max_leaf_nodes < 10 and n_samples >= 1000:
pred_lightgbm = est_lightgbm.predict(X_test)
pred_sklearn = est_sklearn.predict(X_test)
assert np.mean(pred_sklearn == pred_lightgbm) > .89
acc_lightgbm = accuracy_score(y_test, pred_lightgbm)
acc_sklearn = accuracy_score(y_test, pred_sklearn)
np.testing.assert_almost_equal(acc_lightgbm, acc_sklearn, decimal=2)
def common_test_model_hgb_classifier(self, add_nan=False, n_classes=2):
model = HistGradientBoostingClassifier(max_iter=5, max_depth=2)
X, y = make_classification(n_features=10,
n_samples=1000,
n_informative=4,
n_classes=n_classes,
random_state=42)
if add_nan:
rows = numpy.random.randint(0, X.shape[0] - 1, X.shape[0] // 3)
cols = numpy.random.randint(0, X.shape[1] - 1, X.shape[0] // 3)
X[rows, cols] = numpy.nan
X_train, X_test, y_train, _ = train_test_split(X,
y,
test_size=0.5,
random_state=42)
model.fit(X_train, y_train)
if n_classes == 2:
model_onnx = convert_sklearn(
model,
"unused", [("input", FloatTensorType([None, X.shape[1]]))],
options={model.__class__: {
'raw_scores': True
}})
self.assertIsNotNone(model_onnx)
X_test = X_test.astype(numpy.float32)[:5]
dump_data_and_model(
X_test,
model,
model_onnx,
basename="SklearnHGBClassifierRaw%s%d" %
("nan" if add_nan else '', n_classes),
verbose=False,
intermediate_steps=True,
methods=['predict', 'decision_function_binary'],
backend=['python'])
model_onnx = convert_sklearn(
model, "unused", [("input", FloatTensorType([None, X.shape[1]]))])
self.assertIsNotNone(model_onnx)
X_test = X_test.astype(numpy.float32)[:5]
dump_data_and_model(X_test,
model,
model_onnx,
basename="SklearnHGBClassifier%s%d" %
("nan" if add_nan else '', n_classes),
verbose=False)
def test_missing_values_trivial():
# sanity check for missing values support. With only one feature and
# y == isnan(X), the gbdt is supposed to reach perfect accuracy on the
# training set.
n_samples = 100
n_features = 1
rng = np.random.RandomState(0)
X = rng.normal(size=(n_samples, n_features))
mask = rng.binomial(1, 0.5, size=X.shape).astype(bool)
X[mask] = np.nan
y = mask.ravel()
gb = HistGradientBoostingClassifier()
gb.fit(X, y)
assert gb.score(X, y) == pytest.approx(1)
def k_fold_cross_val(Xs, y_var, k=10):
clf = tree.DecisionTreeClassifier()
clf_forest = RandomForestClassifier(n_estimators=10)
clf_boost = HistGradientBoostingClassifier()
num_folds = k
N = Xs.shape[0]
test_size = int(N / num_folds)
test_idxs = np.random.permutation(N)[:num_folds * test_size].reshape(
num_folds, test_size)
total_score = np.asarray([0., 0., 0.])
total_F1_score = np.asarray([0., 0., 0.])
for i in range(num_folds):
print("Iteration " + str(i) + ":")
test_i = Xs.index.isin(test_idxs[i])
df_train, df_test = Xs[~test_i], Xs[test_i]
y_train, y_test = y_var[~test_i], y_var[test_i]
clf = clf.fit(df_train.to_numpy(), y_train.to_numpy().ravel())
score_b = clf.score(df_test.to_numpy(), y_test.to_numpy().ravel())
clf_forest = clf_forest.fit(df_train.to_numpy(),
y_train.to_numpy().ravel())
score_f = clf_forest.score(df_test.to_numpy(),
y_test.to_numpy().ravel())
clf_boost = clf_boost.fit(df_train.to_numpy(),
y_train.to_numpy().ravel())
score_h = clf_boost.score(df_test.to_numpy(),
y_test.to_numpy().ravel())
y_hat = clf.predict(df_test.to_numpy())
f1_b = f1_score(y_test.to_numpy().ravel(), y_hat, average='binary')
print("F1 score (tree):", f1_b)
y_hat = clf_forest.predict(df_test.to_numpy())
f1_f = f1_score(y_test.to_numpy().ravel(), y_hat, average='binary')
print("F1 score (forest):", f1_f)
y_hat = clf_boost.predict(df_test.to_numpy())
f1_boost = f1_score(y_test.to_numpy().ravel(), y_hat, average='binary')
print("F1 score (boost):", f1_boost)
print("Prediction scores for (tree,forest,boost):", score_b, score_f,
score_h)
total_score += np.asarray([score_b, score_f, score_h])
total_F1_score += np.asarray([f1_b, f1_f, f1_boost])
print("Avg. accuracy scores for (tree,forest,boost):",
total_score / num_folds)
print("Avg. F1 scores for (tree,forest,boost):",
total_F1_score / num_folds)
return clf, clf_forest, clf_boost
def test_categorical_encoding_strategies():
# Check native categorical handling vs different encoding strategies. We
# make sure that native encoding needs only 1 split to achieve a perfect
# prediction on a simple dataset. In contrast, OneHotEncoded data needs
# more depth / splits, and treating categories as ordered (just using
# OrdinalEncoder) requires even more depth.
# dataset with one random continuous feature, and one categorical feature
# with values in [0, 5], e.g. from an OrdinalEncoder.
# class == 1 iff categorical value in {0, 2, 4}
rng = np.random.RandomState(0)
n_samples = 10_000
f1 = rng.rand(n_samples)
f2 = rng.randint(6, size=n_samples)
X = np.c_[f1, f2]
y = np.zeros(shape=n_samples)
y[X[:, 1] % 2 == 0] = 1
# make sure dataset is balanced so that the baseline_prediction doesn't
# influence predictions too much with max_iter = 1
assert 0.49 < y.mean() < 0.51
clf_cat = HistGradientBoostingClassifier(
max_iter=1, max_depth=1, categorical_features=[False, True]
)
# Using native categorical encoding, we get perfect predictions with just
# one split
assert cross_val_score(clf_cat, X, y).mean() == 1
# quick sanity check for the bitset: 0, 2, 4 = 2**0 + 2**2 + 2**4 = 21
expected_left_bitset = [21, 0, 0, 0, 0, 0, 0, 0]
left_bitset = clf_cat.fit(X, y)._predictors[0][0].raw_left_cat_bitsets[0]
assert_array_equal(left_bitset, expected_left_bitset)
# Treating categories as ordered, we need more depth / more splits to get
# the same predictions
clf_no_cat = HistGradientBoostingClassifier(
max_iter=1, max_depth=4, categorical_features=None
)
assert cross_val_score(clf_no_cat, X, y).mean() < 0.9
clf_no_cat.set_params(max_depth=5)
assert cross_val_score(clf_no_cat, X, y).mean() == 1
# Using OHEd data, we need less splits than with pure OEd data, but we
# still need more splits than with the native categorical splits
ct = make_column_transformer(
(OneHotEncoder(sparse=False), [1]), remainder="passthrough"
)
X_ohe = ct.fit_transform(X)
clf_no_cat.set_params(max_depth=2)
assert cross_val_score(clf_no_cat, X_ohe, y).mean() < 0.9
clf_no_cat.set_params(max_depth=3)
assert cross_val_score(clf_no_cat, X_ohe, y).mean() == 1
def foo(Xs, Xg, ys, yg):
# For each subset, make splits
kf = KFold(n_splits=n_splits, random_state=0)
# Store ITEs
ite = []
# For each split
for idx1, idx2 in kf.split(Xs):
# Init models
ms = HistGradientBoostingRegressor()
mg = HistGradientBoostingClassifier()
# Train models
ms.fit(Xs[idx1], ys[idx1])
mg.fit(Xg[idx1], yg[idx1])
# Make estimates on test set
ite.append(
AIPW_estimator(ms, mg, Xs[idx2], Xg[idx2], ys[idx2], yg[idx2]))
# Return mean ite and n_employees
return np.concatenate(ite).mean(), len(Xs)
def gradient_boost(train_data, test_data):
train_y = train_data['state']
train_X = train_data.iloc[:, FEATURES_INDICES]
test_y = test_data['state']
test_X = test_data.iloc[:, FEATURES_INDICES]
#search(train_X, train_y)
#search_xgboost(train_X, train_y)
gd = HistGradientBoostingClassifier(loss='auto',
max_bins=200,
max_depth=10,
max_leaf_nodes=35)
#gd = XGBClassifier()
gd.fit(train_X, train_y)
pred_y = gd.predict(test_X)
evaluate(gd, test_X, test_y, pred_y)
def main():
# loading the dataset from sklearn.datasets
df_cancer = load_breast_cancer()
print(df_cancer.keys())
X = df_cancer.data
y = df_cancer.target
print("number of classes are: ", np.unique(y))
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.3,
random_state=0)
# create object of historgradientboosting
hist = HistGradientBoostingClassifier()
# training the model
hist.fit(X_train, y_train)
y_pred = hist.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print("accuracy of the model is: ", accuracy)
clr = classification_report(y_test, y_pred)
print("Classification report is:", clr)
def fit(
self,
params,
X_train,
y_train,
Xy_val,
sample_weight,
n_estimators=None,
seed=None,
):
# Xy_val not used
if seed is not None:
params.update({"random_state": seed})
if n_estimators is not None:
params.update({"max_iter": n_estimators})
clf = HistGradientBoostingClassifier(
**params,
categorical_features=self.categorical_features,
early_stopping="auto")
clf.fit(X_train, y_train, sample_weight=sample_weight)
return clf, None
def test_on_target(rawdata, sitename):
print('------------Testing on %s-----------' % sitename)
target_info = pd.read_csv("target_info.csv")
if sitename in target_info['Site'].values:
target_dict = target_info.set_index('Site').T.to_dict()
sequence = target_dict[sitename]['Sequence']
train_data = rawdata[rawdata[0]!=sequence]
test_data = rawdata[rawdata[0]==sequence]
X_train, y_train = create_Input(train_data)
X_test, y_test = create_Input(test_data)
pmfm = create_pmfm(X_train,y_train)
train_feature = X_train.apply(feature_Encoding, axis = 1, args = (0,1,pmfm)).values
test_feature = X_test.apply(feature_Encoding, axis = 1, args = (0,1,pmfm)).values
clf = HistGradientBoostingClassifier(max_iter=500, learning_rate=0.9, max_depth=6, l2_regularization=100)
train_matrix = np.matrix([train_feature[i] for i in range(train_feature.shape[0])])
test_matrix = np.matrix([test_feature[i] for i in range(test_feature.shape[0])])
clf.fit(train_matrix, y_train)
pred = clf.predict(test_matrix)
evaluate(y_test, pred)
else:
print('ERROR: INCORRECT SITE NAME')
class MyHGBClassifierModel(BaseModel):
def __init__(self, model_params, fit_params: Optional[Dict]):
self.model_params = model_params
self.fit_params = fit_params
if self.fit_params is None:
self.fit_params = {}
def build_model(self):
self.model = HistGradientBoostingClassifier(**self.model_params)
return self.model
def fit(self, train_x, train_y, valid_x=None, valid_y=None):
self.model = self.build_model()
self.model.fit(
train_x, train_y,
**self.fit_params
)
return self.model
def predict(self, est, valid_x):
preds = est.predict_proba(valid_x)[:, 1]
return preds
def test_early_stopping_classification(data, scoring, validation_fraction,
n_iter_no_change, tol):
max_iter = 50
X, y = data
gb = HistGradientBoostingClassifier(
verbose=1, # just for coverage
min_samples_leaf=5, # easier to overfit fast
scoring=scoring,
tol=tol,
validation_fraction=validation_fraction,
max_iter=max_iter,
n_iter_no_change=n_iter_no_change,
random_state=0
)
gb.fit(X, y)
if n_iter_no_change is not None:
assert n_iter_no_change <= gb.n_iter_ < max_iter
else:
assert gb.n_iter_ == max_iter
def test_binning_train_validation_are_separated():
# Make sure training and validation data are binned separately.
# See issue 13926
rng = np.random.RandomState(0)
validation_fraction = .2
gb = HistGradientBoostingClassifier(
n_iter_no_change=5,
validation_fraction=validation_fraction,
random_state=rng
)
gb.fit(X_classification, y_classification)
mapper_training_data = gb.bin_mapper_
# Note that since the data is small there is no subsampling and the
# random_state doesn't matter
mapper_whole_data = _BinMapper(random_state=0)
mapper_whole_data.fit(X_classification)
n_samples = X_classification.shape[0]
assert np.all(mapper_training_data.actual_n_bins_ ==
int((1 - validation_fraction) * n_samples))
assert np.all(mapper_training_data.actual_n_bins_ !=
mapper_whole_data.actual_n_bins_)
def common_test_model_hgb_classifier(self, add_nan=False, n_classes=2):
model = HistGradientBoostingClassifier(max_iter=5, max_depth=2)
X, y = make_classification(n_features=10,
n_samples=1000,
n_informative=4,
n_classes=n_classes,
random_state=42)
if add_nan:
rows = numpy.random.randint(0, X.shape[0] - 1, X.shape[0] // 3)
cols = numpy.random.randint(0, X.shape[1] - 1, X.shape[0] // 3)
X[rows, cols] = numpy.nan
X_train, X_test, y_train, _ = train_test_split(X,
y,
test_size=0.5,
random_state=42)
model.fit(X_train, y_train)
model_onnx = convert_sklearn(
model, "unused", [("input", FloatTensorType([None, X.shape[1]]))])
self.assertIsNotNone(model_onnx)
X_test = X_test.astype(numpy.float32)[:5]
dump_data_and_model(X_test, model, model_onnx, folder=self.folder)
def test_infinite_values_missing_values():
# High level test making sure that inf and nan values are properly handled
# when both are present. This is similar to
# test_split_on_nan_with_infinite_values() in test_grower.py, though we
# cannot check the predicitons for binned values here.
X = np.asarray([-np.inf, 0, 1, np.inf, np.nan]).reshape(-1, 1)
y_isnan = np.isnan(X.ravel())
y_isinf = X.ravel() == np.inf
stump_clf = HistGradientBoostingClassifier(min_samples_leaf=1, max_iter=1,
learning_rate=1, max_depth=2)
assert stump_clf.fit(X, y_isinf).score(X, y_isinf) == 1
assert stump_clf.fit(X, y_isnan).score(X, y_isnan) == 1
def clasificar_HistGradientBoostingClassifier(X, y, df, trainInputs,
trainOutputs, testInputs,
testOutputs, graphname):
print("\n[" + str(graphname) + "]")
scoreArray = np.array([])
clf = HistGradientBoostingClassifier()
scores = cross_val_score(clf, X, y, cv=10)
clf = clf.fit(trainInputs, trainOutputs)
precisionTrain = clf.score(trainInputs, trainOutputs)
precisionTest = clf.score(testInputs, testOutputs)
print("\tCCR train = %.2f%% | CCR test = %.2f%%" %
(precisionTrain * 100, precisionTest * 100))
prediccion_test = clf.predict(testInputs)
print(prediccion_test)
print(testOutputs)
return precisionTest
data_train, target_train = data_train[:subsample], target_train[:subsample]
n_samples, n_features = data_train.shape
print(f"Training set with {n_samples} records with {n_features} features.")
print("Fitting a sklearn model...")
tic = time()
est = HistGradientBoostingClassifier(loss='binary_crossentropy',
learning_rate=lr,
max_iter=n_trees,
max_bins=max_bins,
max_leaf_nodes=n_leaf_nodes,
n_iter_no_change=None,
random_state=0,
verbose=1)
est.fit(data_train, target_train)
toc = time()
predicted_test = est.predict(data_test)
predicted_proba_test = est.predict_proba(data_test)
roc_auc = roc_auc_score(target_test, predicted_proba_test[:, 1])
acc = accuracy_score(target_test, predicted_test)
print(f"done in {toc - tic:.3f}s, ROC AUC: {roc_auc:.4f}, ACC: {acc :.4f}")
if args.lightgbm:
print("Fitting a LightGBM model...")
tic = time()
lightgbm_est = get_equivalent_estimator(est, lib='lightgbm')
lightgbm_est.fit(data_train, target_train)
toc = time()
predicted_test = lightgbm_est.predict(data_test)
predicted_proba_test = lightgbm_est.predict_proba(data_test)
# ############################################################ BaggingClassifier
clf_bc = BaggingClassifier(base_estimator=SVC(),
n_estimators=10,
random_state=0)
clf_bc.fit(x_train, y_train)
bc_pred = clf_bc.predict(x_test)
bc_matrices = evaluate_preds(clf_bc, x_test, y_test, bc_pred)
# ################################################ ExtraTreesClassifier
clf_etc = ExtraTreesClassifier()
clf_etc.fit(x_train, y_train)
etc_pred = clf_etc.predict(x_test)
et_matrices = evaluate_preds(clf_etc, x_test, y_test, etc_pred)
# ############################################################
# ############################################################ HistGradientBoostingClassifier
clf_hgbc = HistGradientBoostingClassifier()
clf_hgbc.fit(x_train, y_train)
hgbc_pred = clf_hgbc.predict(x_test)
hgb_matrices = evaluate_preds(clf_hgbc, x_test, y_test, hgbc_pred)
# ############################################################
# ############################################################ LogisticRegression
clf_lr = LogisticRegression()
clf_lr.fit(x_train, y_train)
clf_pred = clf_lr.predict(x_test)
lr_matrices = evaluate_preds(clf_lr, x_test, y_test, clf_pred)
# ############################################################
# ############################################################ StackingClassifier
clf_sc = StackingClassifier(estimators=estimators,
final_estimator=LogisticRegression())
clf_sc.fit(x_train, y_train)
clf_pred = clf_sc.predict(x_test)
sc_matrices = evaluate_preds(clf_sc, x_test, y_test, clf_pred)
multi_clf.fit(patient_train_data, train_labels.to_numpy()[:, 1:11])
res = 1 / (1 + np.exp(-multi_clf.decision_function(patient_test_data)))
task1_df = pd.DataFrame(data=res, columns=train_labels.columns[1:11])
task1_df.to_csv('subtask1.csv', index=False, header=True)
t2 = time.time()
print('subtask1, time taken: ', t2 - t1)
print(task1_df)
#subtask 2
t1 = time.time()
clf.fit(patient_train_data, train_labels.to_numpy()[:, 11])
res = 1 / (1 + np.exp(-clf.decision_function(patient_test_data)))
task2_df = pd.DataFrame(data=res, columns=[train_labels.columns[11]])
task2_df.to_csv('subtask2.csv', index=False, header=True)
t2 = time.time()
print('subtask2, time taken: ', t2 - t1)
print(task2_df)
#subtask 3
t1 = time.time()
reg = HistGradientBoostingRegressor(random_state=1510)