tdfs 0.0.4

Feature Store client using Teradata Dataframes

Teradata Feature Store Client

• Installation
• Usage

Installation

You can install via pip. The minimum python version required is 3.6+

pip install tdfs

Usage

First, you'll need teradataml context:

from teradataml.context.context import *
import getpass
create_context('mycluster.teradata.com', 'my_user', getpass.getpass())

Next, use this context to get connection and instantiate Feature Store client:

from tdfs import FeatureStore
fs = FeatureStore(get_connection(), metadata_table = 'my_database.fs_metadata')

It will create metadata table if it doesn't exist.

Register feature groups

fs.register_feature_group(entity = 'patient', table = 'dm.cust_profile', entity_key_column = 'cust_id', features = {'Age':'Age'}, date_column = 'snapshot_dt')

Discover registered features

fs.list_entities()

fs.list_features('patient')

fs.get_feature_details('patient', 'Age')

Get feature set for training/evaluation/scoring

train_df = fs.get_featureset_df('patient','2021-07-12',['Age','DiPedFunc', 'TwoHourSerIns'])

train_pdf = train_df.to_pandas()