def _get_or_create_model(self, ticker):
log = logging.getLogger()
model_filename = self.get_model_filename_from_ticker(ticker)
model = get_model_from_bucket(model_filename, self.get_bucket_name())
if model is None:
log.warning(f'training model for {ticker}')
model = self._model_creator.fit(ticker)
with open(model_filename, 'wb') as f:
joblib.dump(model, f)
upload_file_to_bucket(model_filename, self.get_bucket_name())
return model
def _get_or_create_model(self, ticker):
log = logging.getLogger()
#Sets the filename of the model as ticker.pkl
model_filename = self.get_model_filename_from_ticker(ticker)
#Sets model to the saved model or none if it doesn't exist
model = get_model_from_bucket(model_filename, self.get_bucket_name())
if model is None:
log.warning(f'training model for {ticker}')
#Here is where we fit the model if it doesn't already exist
model = self._model_creator.fit()
with open(model_filename, 'wb') as f:
joblib.dump(model, f)
#Where we upload and save the new model
upload_file_to_bucket(model_filename, self.get_bucket_name())
return model
def get_or_build_model(X_train, y_train, model_name, rebuild_model=False):
"""
Get or build the model of the specific sector name with all the data of all tickers
"""
model_filename = f'{model_name}.pkl'
model = get_model_from_bucket(model_filename, bucket_name)
if rebuild_model or not model:
model = RandomForestClassifier(n_estimators=para_estimator,
max_depth=para_max_depth,
min_samples_leaf=para_min_leaf,
oob_score=True,
random_state=para_random_state)
model.fit(X_train, y_train)
with open(model_filename, 'wb') as f:
joblib.dump(model, f)
upload_file_to_bucket(model_filename, bucket_name)
return model
