def scale_svd_rf_pipe():
from h2o.transforms.decomposition import H2OSVD
print("Importing USArrests.csv data...")
arrests = h2o.upload_file(
pyunit_utils.locate("smalldata/pca_test/USArrests.csv"))
# build transformation pipeline using sklearn's Pipeline and H2OSVD
pipe = Pipeline([("standardize", H2OScaler()), ("svd", H2OSVD()),
("rf", H2ORandomForestEstimator())])
params = {
"standardize__center": [True, False],
"standardize__scale": [True, False],
"svd__nv": [2, 3],
"rf__ntrees": randint(50, 60),
"rf__max_depth": randint(4, 8),
"rf__min_rows": randint(5, 10),
"svd__transform": ["none", "standardize"],
}
custom_cv = H2OKFold(arrests, n_folds=5, seed=42)
random_search = RandomizedSearchCV(pipe,
params,
n_iter=5,
scoring=make_scorer(h2o_r2_score),
cv=custom_cv,
random_state=42,
n_jobs=1)
random_search.fit(arrests[1:], arrests[0])
print(random_search.best_estimator_)
def scale_pca_rf_pipe_new_import():
from h2o.estimators.pca import H2OPrincipalComponentAnalysisEstimator
iris = h2o.import_file(path=pyunit_utils.locate("smalldata/iris/iris_wheader.csv"))
# build transformation pipeline using sklearn's Pipeline and H2O transforms
pipe = Pipeline([
("standardize", H2OScaler()),
("pca", H2OPrincipalComponentAnalysisEstimator().init_for_pipeline()),
("rf", H2ORandomForestEstimator())
])
params = {"standardize__center": [True, False], # Parameters to test
"standardize__scale": [True, False],
"pca__k": randint(2, iris[1:].shape[1]),
"rf__ntrees": randint(50,60),
"rf__max_depth": randint(4,8),
"rf__min_rows": randint(5,10),
"pca__transform": ["none", "standardize"],
}
custom_cv = H2OKFold(iris, n_folds=5, seed=42)
random_search = RandomizedSearchCV(pipe,
params,
n_iter=5,
scoring=make_scorer(h2o_r2_score),
cv=custom_cv,
random_state=42,
n_jobs=1)
random_search.fit(iris[1:],iris[0])
print(random_search.best_estimator_)
def test_h2o_only_pipeline_with_h2o_frames():
pipeline = Pipeline([('svd', H2OSVD(seed=seed)),
('estimator',
H2OGradientBoostingClassifier(seed=seed))])
params = dict(
svd__nv=[2, 3],
svd__transform=['DESCALE', 'DEMEAN', 'NONE'],
estimator__ntrees=[5, 10],
estimator__max_depth=[1, 2, 3],
estimator__learn_rate=[0.1, 0.2],
)
search = RandomizedSearchCV(
pipeline,
params,
n_iter=5,
random_state=seed,
n_jobs=1, # fails with parallel jobs
)
data = _get_data(format='h2o', n_classes=3)
assert isinstance(data.X_train, h2o.H2OFrame)
search.set_params(
scoring=make_scorer(_h2o_accuracy),
cv=H2OKFold(data.X_train, n_folds=3, seed=seed),
)
search.fit(data.X_train, data.y_train)
preds = search.predict(data.X_test)
assert isinstance(preds, h2o.H2OFrame)
assert preds.dim == [len(data.X_test), 1]
probs = search.predict_proba(data.X_test)
assert probs.dim == [len(data.X_test), 3]
assert np.allclose(np.sum(probs.as_data_frame().values, axis=1),
1.), "`predict_proba` didn't return probabilities"
score = search.score(data.X_test, data.y_test)
assert isinstance(score, float)
skl_score = accuracy_score(data.y_test.as_data_frame().values,
preds.as_data_frame().values)
assert abs(score - skl_score) < 1e-6, "score={}, skl_score={}".format(
score, skl_score)
scores['h2o_only_pipeline_with_h2o_frame'] = score
def scale_pca_rf_pipe():
from h2o.transforms.preprocessing import H2OScaler
from h2o.estimators.pca import H2OPrincipalComponentAnalysisEstimator
from h2o.estimators.random_forest import H2ORandomForestEstimator
from sklearn.pipeline import Pipeline
from sklearn.grid_search import RandomizedSearchCV
from h2o.cross_validation import H2OKFold
from h2o.model.regression import h2o_r2_score
from sklearn.metrics.scorer import make_scorer
from scipy.stats import randint
iris = h2o.import_file(
path=pyunit_utils.locate("smalldata/iris/iris_wheader.csv"))
# build transformation pipeline using sklearn's Pipeline and H2O transforms
pipe = Pipeline([("standardize", H2OScaler()),
("pca", H2OPrincipalComponentAnalysisEstimator()),
("rf", H2ORandomForestEstimator())])
params = {
"standardize__center": [True, False], # Parameters to test
"standardize__scale": [True, False],
"pca__k": randint(2, iris[1:].shape[1]),
"rf__ntrees": randint(50, 60),
"rf__max_depth": randint(4, 8),
"rf__min_rows": randint(5, 10),
}
custom_cv = H2OKFold(iris, n_folds=5, seed=42)
random_search = RandomizedSearchCV(pipe,
params,
n_iter=5,
scoring=make_scorer(h2o_r2_score),
cv=custom_cv,
random_state=42,
n_jobs=1)
random_search.fit(iris[1:], iris[0])
print(random_search.best_estimator_)
("pca", H2OPCA(k=2)),
("gbm", H2OGradientBoostingEstimator(distribution="multinomial"))])
pipeline.fit(iris_df[:4],iris_df[4])
# Random CV using H2O and Scikit-learn
from sklearn.grid_search import RandomizedSearchCV
from h2o.cross_validation import H2OKFold
from h2o.model.regression import h2o_r2_score
from sklearn.metrics.scorer import make_scorer
params = {"standardize__center": [True, False], # Parameters to test
"standardize__scale": [True, False],
"pca__k": [2,3],
"gbm__ntrees": [10,20],
"gbm__max_depth": [1,2,3],
"gbm__learn_rate": [0.1,0.2]}
custom_cv = H2OKFold(iris_df, n_folds=5, seed=42)
pipeline = Pipeline([("standardize", H2OScaler()),
("pca", H2OPCA(k=2)),
("gbm", H2OGradientBoostingEstimator(distribution="gaussian"))])
random_search = RandomizedSearchCV(pipeline, params,
n_iter=5,
scoring=make_scorer(h2o_r2_score),
cv=custom_cv,
random_state=42,
n_jobs=1)
random_search.fit(iris_df[1:], iris_df[0])
print random_search.best_estimator_
('groupby_count', AddGroupByCount())])),
('Add_ip_address_num',
Pipeline([('extract', ColumnExtractor(['user_id', 'ip_address'])),
('groupby_count', AddGroupByCount())])),
('numerics',
Pipeline([('extract', ColumnExtractor(NUM_FEATS)),
('zero_fill', ZeroFillTransformer()),
('log', Log1pTransformer())]))
]))
])
##############################
# Modeling + Tuning
##############################
from h2o.cross_validation import H2OKFold
dataset = pd.concat([X_train, y_train], axis=1)
cv = H2OKFold(dataset, n_folds=5, seed=42)
# H2O approach
("H2OCreator", H2OFrameCreator()),
# ('standardize', H2OScaler()),
# ('pca', H2OPCA()),
('rf', H2ORandomForestEstimator(ntrees=20))
# something new to try
# from scipy.stats import randint
# params = {
# # "standardize__center": [True, False],
# # "standardize__scale": [True, False],
# "pca__k": 2,
# # randint(2, X_train[1:].shape[1]),
# "rf__ntrees": 20
# # randint(50,60),