def test_selective_imputer_bad_strategies():
# raises for a bad strategy string
imputer = SelectiveImputer(strategy="bad strategy")
assert_raises(ValueError, imputer.fit, X)
# raises for a dim mismatch in cols and strategy
imputer = SelectiveImputer(cols=['a'], strategy=['mean', 'mean'])
assert_raises(ValueError, imputer.fit, X)
# test type error for bad strategy
imputer = SelectiveImputer(strategy=1)
assert_raises(TypeError, imputer.fit, X)
# test dict input that does not match dim-wise
imputer = SelectiveImputer(cols=['a'],
strategy={
'a': 'mean',
'b': 'median'
})
assert_raises(ValueError, imputer.fit, X)
# test a dict input with bad columns breaks
imputer = SelectiveImputer(strategy={'a': 'mean', 'D': 'median'})
assert_raises(ValueError, imputer.fit, X)
def test_complex_grid_search():
# build a pipeline
pipe = Pipeline([
('dropper', FeatureFilter()), # won't drop any
('collinearity', MultiCorrFilter(threshold=0.85)),
('imputer', SelectiveImputer()), # pass through since all full
('scaler', SelectiveMaxAbsScaler()),
('boxcox', BoxCoxTransformer(suppress_warnings=True)),
('nzv', NearZeroVarianceFilter()),
('pca', SelectivePCA(n_components=0.9)),
('custom', make_transformer(subtract_k, k=1)),
('model', RandomForestClassifier(n_jobs=1))
])
# let's define a set of hyper-parameters over which to search
hp = {
'collinearity__threshold': uniform(loc=.8, scale=.15),
'collinearity__method': ['pearson', 'kendall', 'spearman'],
'pca__n_components': uniform(loc=.75, scale=.2),
'pca__whiten': [True, False],
'custom__k': [1, 2, 3],
'custom__func': [subtract_k, add_k],
'model__n_estimators': randint(5, 10),
'model__max_depth': randint(2, 5),
'model__min_samples_leaf': randint(1, 5),
'model__max_features': uniform(loc=.5, scale=.5),
'model__max_leaf_nodes': randint(10, 15)
}
# define the gridsearch
search = RandomizedSearchCV(
pipe,
hp,
n_iter=2, # just to test it even works
scoring='accuracy',
cv=cv,
random_state=42,
# in parallel so we are testing pickling of the classes
n_jobs=2)
# fit the search
search.fit(X_train, y_train)
# Show we can profile the best estimator
profile_estimator(search.best_estimator_)
# Assert that it's persistable
assert_persistable(pipe, "location.pkl", X_train, y_train)
def test_complex_grid_search():
# build a pipeline
pipe = Pipeline([
('dropper', FeatureFilter()), # won't drop any
('collinearity', MultiCorrFilter(threshold=0.85)),
('imputer', SelectiveImputer()), # pass through since all full
('scaler', SelectiveScaler()),
('boxcox', BoxCoxTransformer()),
('nzv', NearZeroVarianceFilter()),
('pca', SelectivePCA(n_components=0.9)),
('model', RandomForestClassifier(n_jobs=1))
])
# let's define a set of hyper-parameters over which to search
hp = {
'collinearity__threshold': uniform(loc=.8, scale=.15),
'collinearity__method': ['pearson', 'kendall', 'spearman'],
'scaler__scaler': [None, RobustScaler()],
'pca__n_components': uniform(loc=.75, scale=.2),
'pca__whiten': [True, False],
'model__n_estimators': randint(5, 10),
'model__max_depth': randint(2, 5),
'model__min_samples_leaf': randint(1, 5),
'model__max_features': uniform(loc=.5, scale=.5),
'model__max_leaf_nodes': randint(10, 15)
}
# define the gridsearch
search = RandomizedSearchCV(
pipe,
hp,
n_iter=2, # just to test it even works
scoring='accuracy',
cv=cv,
random_state=42)
# fit the search
search.fit(X_train, y_train)