def test_StackingClassifier_drop_last_proba():
np.random.seed(123)
lr1 = LogisticRegression(solver='liblinear',
multi_class='ovr')
sclf1 = StackingCVClassifier(classifiers=[lr1, lr1],
use_probas=True,
drop_last_proba=False,
meta_classifier=lr1)
sclf1.fit(X_iris, y_iris)
r1 = sclf1.predict_meta_features(X_iris[:2])
assert r1.shape == (2, 6)
sclf2 = StackingCVClassifier(classifiers=[lr1, lr1],
use_probas=True,
drop_last_proba=True,
meta_classifier=lr1)
sclf2.fit(X_iris, y_iris)
r2 = sclf2.predict_meta_features(X_iris[:2])
assert r2.shape == (2, 4), r2.shape
sclf3 = StackingCVClassifier(classifiers=[lr1, lr1],
use_probas=True,
drop_last_proba=True,
meta_classifier=lr1)
sclf3.fit(X_iris[0:100], y_iris[0:100]) # only 2 classes
r3 = sclf3.predict_meta_features(X_iris[:2])
assert r3.shape == (2, 2), r3.shape
def test_StackingClassifier_drop_last_proba():
np.random.seed(123)
lr1 = LogisticRegression(solver='liblinear', multi_class='ovr')
sclf1 = StackingCVClassifier(classifiers=[lr1, lr1],
use_probas=True,
drop_last_proba=False,
meta_classifier=lr1)
sclf1.fit(X_iris, y_iris)
r1 = sclf1.predict_meta_features(X_iris[:2])
assert r1.shape == (2, 6)
sclf2 = StackingCVClassifier(classifiers=[lr1, lr1],
use_probas=True,
drop_last_proba=True,
meta_classifier=lr1)
sclf2.fit(X_iris, y_iris)
r2 = sclf2.predict_meta_features(X_iris[:2])
assert r2.shape == (2, 4), r2.shape
sclf3 = StackingCVClassifier(classifiers=[lr1, lr1],
use_probas=True,
drop_last_proba=True,
meta_classifier=lr1)
sclf3.fit(X_iris[0:100], y_iris[0:100]) # only 2 classes
r3 = sclf3.predict_meta_features(X_iris[:2])
assert r3.shape == (2, 2), r3.shape
class Feature(object):
def __init__(self, *, input_path, output_path):
self.__fill_value_transformer_array = importlib.import_module(
"20180603.StackingReallyMax.FillValueTransformerArray")
self.__indicate_feature_transformer_array = importlib.import_module(
"20180603.StackingReallyMax.IndicateFeatureTransformerArray")
self.__input_path = input_path
self.__output_path = output_path
# data prepare
self.__train, self.__test = [None for _ in range(2)]
self.__train_feature, self.__train_label = [None for _ in range(2)]
self.__test_feature = None
# model fit
self.__rfc = None
self.__etc = None
self.__gbc = None
self.__xbc = None
self.__lr_l1 = None
self.__lr_l2 = None
self.__net = None
self.__knc = None
self.__rfc_pl = None
self.__etc_pl = None
self.__gbc_pl = None
self.__xbc_pl = None
self.__lr_l1_pl = None
self.__lr_l2_pl = None
self.__net_pl = None
self.__knc_pl = None
self.__clf = None
self.__oof_train = None
self.__oof_test = None
# model_feature_output
self.__feature_train = None
self.__feature_test = None
def data_prepare(self):
self.__train = pd.read_csv(
os.path.join(self.__input_path, "first_layer_train.csv"))
self.__test = pd.read_csv(
os.path.join(self.__input_path, "first_layer_test.csv"))
self.__train_label = self.__train["TARGET"]
self.__train_feature = self.__train.drop(["TARGET"], axis=1)
self.__test_feature = self.__test
del self.__train, self.__test
gc.collect()
def model_fit(self):
self.__rfc = RandomForestClassifier(n_jobs=-1)
self.__etc = ExtraTreesClassifier(n_jobs=-1)
self.__gbc = GradientBoostingClassifier()
self.__xbc = XGBClassifier(n_jobs=-1)
self.__lr_l1 = LogisticRegression(penalty="l1")
self.__lr_l2 = LogisticRegression(penalty="l2")
self.__net = MLPClassifier()
self.__knc = KNeighborsClassifier(n_jobs=-1)
self.__rfc_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=-999.0)), ("Clf", self.__rfc)])
self.__etc_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=-999.0)), ("Clf", self.__etc)])
self.__gbc_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=-999.0)), ("Clf", self.__gbc)])
self.__xbc_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=-999.0)), ("Clf", self.__xbc)])
self.__lr_l1_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=0)), ("Clf", self.__lr_l1)])
self.__lr_l2_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=0)), ("Clf", self.__lr_l2)])
self.__net_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=0)), ("Clf", self.__net)])
self.__knc_pl = Pipeline(
steps=[("AddIndicator",
self.__indicate_feature_transformer_array.
IndicateFeatureTransformerArray(columns=[9, 10, 11])),
("FillNa",
self.__fill_value_transformer_array.
FillValueTransformerArray(
filling_values=0)), ("Clf", self.__knc)])
self.__clf = StackingCVClassifier(classifiers=[
self.__rfc_pl, self.__etc_pl, self.__gbc_pl, self.__xbc_pl,
self.__lr_l1_pl, self.__lr_l2_pl, self.__net_pl, self.__knc_pl
],
meta_classifier=self.__lr_l1_pl,
use_probas=True,
cv=2,
store_train_meta_features=True,
verbose=True)
self.__clf.fit(self.__train_feature.values, self.__train_label.values)
self.__oof_train = self.__clf.train_meta_features_
self.__oof_test = self.__clf.predict_meta_features(
self.__test_feature.values)
self.__oof_train = self.__oof_train[:, [
i for i in range(2 * len(self.__clf.classifiers)) if i % 2 != 0
]]
self.__oof_test = self.__oof_test[:, [
i for i in range(2 * len(self.__clf.classifiers)) if i % 2 != 0
]]
self.__oof_train = pd.DataFrame(self.__oof_train,
columns=[
"rf_2", "et_2", "gb_2", "xg_2",
"lr_l1_2", "lr_l2_2", "net_2",
"knc_2"
])
self.__oof_test = pd.DataFrame(self.__oof_test,
columns=[
"rf_2", "et_2", "gb_2", "xg_2",
"lr_l1_2", "lr_l2_2", "net_2",
"knc_2"
])
def model_feature_output(self):
self.__feature_train = pd.concat(
[self.__train_feature, self.__oof_train], axis=1)
self.__feature_test = pd.concat([self.__test_feature, self.__oof_test],
axis=1)
self.__feature_train.to_csv(os.path.join(self.__output_path,
"feature_train_res.csv"),
index=False)
self.__feature_test.to_csv(os.path.join(self.__output_path,
"feature_test_res.csv"),
index=False)
