class FeatureExtractorClassifier(object):
"""
Difference with the FeatureExtractorClassifier from ramp-workflow:
`test_submission` wraps the y_proba in a DataFrame with the original
index.
"""
def __init__(self):
self.element_names = ['feature_extractor', 'classifier']
self.feature_extractor_workflow = FeatureExtractor(
[self.element_names[0]])
self.classifier_workflow = Classifier([self.element_names[1]])
def train_submission(self, module_path, X_df, y_array, train_is=None):
if train_is is None:
train_is = slice(None, None, None)
fe = self.feature_extractor_workflow.train_submission(
module_path, X_df, y_array, train_is)
X_train_array = self.feature_extractor_workflow.test_submission(
fe, X_df.iloc[train_is])
clf = self.classifier_workflow.train_submission(
module_path, X_train_array, y_array[train_is])
return fe, clf
def test_submission(self, trained_model, X_df):
fe, clf = trained_model
X_test_array = self.feature_extractor_workflow.test_submission(
fe, X_df)
y_proba = self.classifier_workflow.test_submission(clf, X_test_array)
arr = X_df.index.values.astype('datetime64[m]').astype(int)
y = np.hstack((arr[:, np.newaxis], y_proba))
return y
def __init__(self):
self.element_names = ['feature_extractor', 'classifier']
self.feature_extractor_workflow = FeatureExtractor(
[self.element_names[0]])
self.classifier_workflow = Classifier([self.element_names[1]])