class ClassificationPipeline(Pipeline):
"""
Classification pipeline
"""
def __init__(self, **pipeline_constructor_params):
self.estimator = AutoSklearnClassifier(**pipeline_constructor_params)
def fit(self, x, y):
self.estimator.fit(x, y)
def run(self, x):
return self.estimator.predict(x)
def as_json(self):
pipeline = self.estimator.get_models_with_weights()[0][1]
return pipeline.config.get_dictionary()
def fit_automl(self, run_time):
"""Runs auto-sklearn on the uploaded data and prints results.
Side effects:
- Enables upload_widget
Args:
run_time (int): The run time for auto-sklearn in seconds.
Returns:
automl (AutoSklearnClassifier): fitted auto-sklearn model.
"""
automl_args = {}
automl_args['time_left_for_this_task'] = run_time
# TODO functionality to load this from Mongo
automl_args['metadata_directory'] = ".metalearning/metalearning_files/"
#automl_args['metadata_directory'] = "../metalearning/metalearning_files/"
automl = AutoSklearnClassifier(**automl_args)
thread = threading.Thread(target=self.update_progress,
args=(self.progress_widget, ))
thread.start()
# always load a copy of the latest dataset
cur_data = self.data[-1].copy()
y = cur_data.pop(0)
X, feat_types, _ = model_utils.process_feat_types(cur_data)
X_train = X.iloc[self.train_idxs]
y_train = y.iloc[self.train_idxs]
X_test = X.iloc[self.test_idxs]
y_test = y.iloc[self.test_idxs]
with warnings.catch_warnings():
warnings.simplefilter("ignore")
with HiddenPrints():
automl.fit(X_train, y_train, feat_type=feat_types)
# Automl has finished fitting:
self.models.append(copy.deepcopy(automl))
with self.event_output_widget:
print("FITTING COMPLETED WITH FITTING TIME PARAMETER AS ",
int(run_time / 60), " MINUTES")
with self.metrics_output_widget:
y_train_hat = automl.predict(X_train)
train_accuracy_score = metrics.accuracy_score(y_train, y_train_hat)
y_test_hat = automl.predict(X_test)
test_accuracy_score = metrics.accuracy_score(y_test, y_test_hat)
thresholdout_score = model_utils.thresholdout(
train_accuracy_score, test_accuracy_score)
output_str = "Run {}: train acc: {:.4}, noised test acc: {:.4}\n".format(
self.queries, train_accuracy_score, thresholdout_score)
print(output_str)
with self.model_output_widget:
print("MODELS:")
print(automl.get_models_with_weights())
if self.textbox_upload:
self.upload_button.disabled = False
self.upload_text.disabled = False
else:
self.upload_widget.disabled = False
if self.queries == self.budget_widget.value:
self.on_budget_completion()
return automl
data = dataframe.values
X, y = data[:, :-1], data[:, -1]
# minimally prepare dataset
X = X.astype('float32')
y = LabelEncoder().fit_transform(y.astype('str'))
# split into train and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=1)
# define search
model = AutoSklearnClassifier(time_left_for_this_task=10*60, per_run_time_limit=45, n_jobs=6)
# perform the search
model.fit(X_train, y_train)
# summarize
print(model.sprint_statistics())
# get model and weights
model_weights = model.get_models_with_weights()
for model_weight in model_weights:
print(model_weight)
print("Show models")
models_def = model.show_models()
print(models_def)
# evaluate best model
y_hat = model.predict(X_test)
acc = accuracy_score(y_test, y_hat)
print("Test Dataset Accuracy: %.3f" % acc)
