Previous topic

DataContext

Next topic

Features

This Page

Configurations

class ramp.configuration.Configuration(target=None, features=None, model=None, metrics=None, reporters=None, column_subset=None, prediction=None, predictions_name=None, actual=None)

Defines a specific data analysis model, including features, estimator and target metric. Can be stored (pickled) and retrieved.

Args

target: Feature or basestring specifying the target (“y”) variable of the analysis.

features: an iterable of Features <Feature> to be used by the estimator in the analysis.

model: an estimator instance compatible with sklearn estimator conventions. (has fit(x, y) and predict(y) methods).

metrics: an iterable of evaluation `Metric`s used to score predictions.

reporters: an iterable of Reporter objects

prediction: a Feature transformation of the special predictions_name column used to post-process predictions prior to metric scoring.

predictions_name: a unique string used as a column identifier for model predictions. Must be unique among all feature names: eg ‘$logreg_predictions$’

actual: a Feature. Used if target represents a transformation that is NOT the actual target “y” values. Used in conjuction with
prediction to allow model training, predictions and scoring to operate on different values.
class ramp.configuration.ConfigFactory(base_config, **kwargs)

Provides an iterator over passed in configuration values, allowing for easy exploration of models.

Args

base_config: The base Configuration to augment

kwargs: Can be any keyword accepted by Configuration. Values should be iterables.