Components: transform

autogluon.eda.analysis.transform

ApplyFeatureGenerator

This wrapper provides transformed features to all children shadowing outer datasets with the updated one after application of FeatureGenerator.

ApplyFeatureGenerator

class autogluon.eda.analysis.transform.ApplyFeatureGenerator(parent: Optional[autogluon.eda.analysis.base.AbstractAnalysis] = None, children: Optional[List[autogluon.eda.analysis.base.AbstractAnalysis]] = None, state: Optional[autogluon.eda.state.AnalysisState] = None, category_to_numbers: bool = False, feature_generator: Optional[autogluon.features.generators.abstract.AbstractFeatureGenerator] = None, **kwargs)[source]

This wrapper provides transformed features to all children shadowing outer datasets with the updated one after application of FeatureGenerator.

Parameters
category_to_numbers: bool, default = False

if True’, then transform `category variables into their codes. This is useful when wrapped analyses expect numeric values

feature_generator: Optional[AbstractFeatureGenerator], default = None

feature generator to use for the transformation. If None is provided then AutoMLPipelineFeatureGenerator is applied.

parent: Optional[AbstractAnalysis], default = None

parent Analysis

children: Optional[List[AbstractAnalysis]], default None

wrapped analyses; these will receive sampled args during fit call

kwargs
See alsofunc:autogluon.features.AbstractFeatureGenerator

Examples

>>> from autogluon.eda.analysis.base import BaseAnalysis, Namespace
>>> import pandas as pd
>>> import numpy as np
>>> df_train = pd.DataFrame(...)
>>> df_test = pd.DataFrame(...)
>>>
>>> analysis = BaseAnalysis(train_data=df_train, test_data=df_test, label='D', children=[
>>>     Namespace(namespace='feature_generator_numbers', children=[
>>>         ApplyFeatureGenerator(category_to_numbers=True, children=[
>>>             # SomeAnalysis()  # This analysis will be called with transformed `train_data` and `test_data`
>>>         ])
>>>     ]),
>>>     # SomeAnalysis()  # This analysis will be called with the original features
>>> ])