Skip to main content

Featuretools Transformer for Scikit-Learn Pipeline use.

Project description

featuretools-sklearn-transformer

CircleCI codecov

Featuretools' DFS as a scikit-learn transformer

Install

pip install featuretools_sklearn_transformer

Use

To use the transformer in a pipeline, initialize an instance of the transformer by passing in the parameters you would like to use for calculating features. To fit the model and generate features for the training data, pass in an entityset or list of entities and relationships containing only the relevant training data as the X input, along with the training targets as the y input. To generate a feature matrix from test data, pass in an entityset containing only the relevant test data as the X input.

The input supplied for X can take several formats:

  • To use a Featuretools EntitySet without cutoff times, simply pass in the EntitySet
  • To use a Featuretools EntitySet with a cutoff times DataFrame, pass in a tuple of the form (EntitySet, cutoff_time_df)
  • To use a list of Entities and Relationships without cutoff times, pass a tuple of the form (entities, relationships)
  • To use a list of Entities and Relationships with a cutoff times DataFrame, pass a tuple of the form ((entities, relationships), cutoff_time_df)

Note that because this transformer requires a Featuretools EntitySet or Entities and relationships as input, it does not currently work with certain methods such as sklearn.model_selection.cross_val_score or sklearn.model_selection.GridSearchCV which expect the X values to be an iterable which can be split by the method.

The example below shows how to use the transformer with an EntitySet, both with and without a cutoff time DataFrame.

import featuretools as ft
import pandas as pd

from featuretools.wrappers import DFSTransformer
from sklearn.pipeline import Pipeline
from sklearn.ensemble import ExtraTreesClassifier

# Get example data
train_es = ft.demo.load_mock_customer(return_entityset=True, n_customers=3)
test_es = ft.demo.load_mock_customer(return_entityset=True, n_customers=2)
y = [True, False, True]

# Build pipeline
pipeline = Pipeline(steps=[
    ('ft', DFSTransformer(target_entity="customers",
                          max_features=2)),
    ('et', ExtraTreesClassifier(n_estimators=100))
])

# Fit and predict
pipeline.fit(X=train_es, y=y) # fit on customers in training entityset
pipeline.predict_proba(test_es) # predict probability of each class on test entityset
pipeline.predict(test_es) # predict on test entityset

# Same as above, but using cutoff times
train_ct = pd.DataFrame()
train_ct['customer_id'] = [1, 2, 3]
train_ct['time'] = pd.to_datetime(['2014-1-1 04:00',
                                   '2014-1-2 17:20',
                                   '2014-1-4 09:53'])

pipeline.fit(X=(train_es, train_ct), y=y)

test_ct = pd.DataFrame()
test_ct['customer_id'] = [1, 2]
test_ct['time'] = pd.to_datetime(['2014-1-4 13:48',
                                  '2014-1-5 15:32'])
pipeline.predict_proba((test_es, test_ct))
pipeline.predict((test_es, test_ct))

Built at Alteryx Innovation Labs

Alteryx Innovation Labs

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for featuretools-sklearn-transformer, version 0.2.0
Filename, size File type Python version Upload date Hashes
Filename, size featuretools_sklearn_transformer-0.2.0-py3-none-any.whl (9.1 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size featuretools_sklearn_transformer-0.2.0.tar.gz (8.0 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page