def predict_to_catalyst_sam(self, dataframe, server, database, table, schema=None, predicted_column_name=None):
"""
Given a dataframe you want predictions on, make predictions and save them to a catalyst-specific EDW table.
Args:
dataframe (pandas.core.frame.DataFrame): Raw prediction dataframe
server (str): the target server name
database (str): the database name
table (str): the destination table name
schema (str): the optional schema
predicted_column_name (str): optional predicted column name (defaults to PredictedProbNBR or
PredictedValueNBR)
"""
# Make predictions in specific format
sam_df = self.create_catalyst_dataframe(dataframe)
# Rename prediction column to default based on model type or given one
if predicted_column_name is None:
if self.is_classification:
predicted_column_name = 'PredictedProbNBR'
elif self.is_regression:
predicted_column_name = 'PredictedValueNBR'
sam_df.rename(columns={'Prediction': predicted_column_name}, inplace=True)
try:
engine = hcai_db.build_mssql_engine_using_trusted_connections(server, database)
healthcareai.common.database_writers.write_to_db_agnostic(engine, table, sam_df, schema=schema)
except HealthcareAIError as hcaie:
# Run validation and alert user
hcai_dbval.validate_catalyst_prediction_sam_connection(server, table, self.grain_column, self.prediction_column)
raise HealthcareAIError(hcaie.message)
def predict_to_catalyst_sam(self, dataframe, server, database, table, schema=None, predicted_column_name=None):
"""
Given a dataframe you want predictions on, make predictions and save them to a catalyst-specific EDW table.
Args:
dataframe (pandas.core.frame.DataFrame): Raw prediction dataframe
server (str): the target server name
database (str): the database name
table (str): the destination table name
schema (str): the optional schema
predicted_column_name (str): optional predicted column name (defaults to PredictedProbNBR or
PredictedValueNBR)
"""
# Make predictions in specific format
sam_df = self.create_catalyst_dataframe(dataframe)
# Rename prediction column to default based on model type or given one
if predicted_column_name is None:
if self.is_classification:
predicted_column_name = 'PredictedProbNBR'
elif self.is_regression:
predicted_column_name = 'PredictedValueNBR'
sam_df.rename(columns={'Prediction': predicted_column_name}, inplace=True)
try:
engine = hcai_db.build_mssql_engine_using_trusted_connections(server, database)
healthcareai.common.database_writers.write_to_db_agnostic(engine, table, sam_df, schema=schema)
except HealthcareAIError as hcaie:
# Run validation and alert user
hcai_dbval.validate_catalyst_prediction_sam_connection(server, table, self.grain_column, self.prediction_column)
raise HealthcareAIError(hcaie.message)
def test_should_succeed(self):
"""Note this should only run on testing platforms w/ access to a MSSQL server on localhost"""
# TODO clarify EXACTLY what this is testing
# Gin up new fake test database and table
temp_table_name = 'fake_test_table-{}-{}'.format(
uuid.uuid4(), uuid.uuid4())
temp_db_name = 'fake_test_db-{}'.format(uuid.uuid4())
server = 'localhost'
_create_mssql_database(server, temp_db_name)
full_table_path = '[{}].[dbo].[{}]'.format(temp_db_name,
temp_table_name)
_build_deploy_mssql_tables(server, full_table_path)
# Run the test
is_valid = validate_catalyst_prediction_sam_connection(
server, '{}'.format(full_table_path), 'PatientEncounterID',
'[PredictedValueNBR]')
self.assertTrue(is_valid)
# Clean up the mess
_destroy_database(server, temp_db_name)
def test_should_succeed(self):
"""Note this should only run on testing platforms w/ access to a MSSQL server on localhost"""
# TODO clarify EXACTLY what this is testing
# Gin up new fake test database and table
temp_table_name = 'fake_test_table-{}-{}'.format(uuid.uuid4(), uuid.uuid4())
temp_db_name = 'fake_test_db-{}'.format(uuid.uuid4())
server = 'localhost'
_create_mssql_database(server, temp_db_name)
full_table_path = '[{}].[dbo].[{}]'.format(temp_db_name, temp_table_name)
_build_deploy_mssql_tables(server, full_table_path)
# Run the test
is_valid = validate_catalyst_prediction_sam_connection(
server,
'{}'.format(full_table_path),
'PatientEncounterID',
'[PredictedValueNBR]')
self.assertTrue(is_valid)
# Clean up the mess
_destroy_database(server, temp_db_name)