featureInfo-class: Feature information object.
In familiar: End-to-End Automated Machine Learning and Model Evaluation

featureInfo-class

R Documentation

Feature information object.

Description

A featureInfo object contains information for a single feature. This information is used to check data prospectively for consistency and for data preparation. These objects are, for instance, attached to a familiarModel object so that data can be pre-processed in the same way as the development data.

Slots

name: Name of the feature, which by default is the column name of the feature.
set_descriptor: Character string describing the set to which the feature belongs. Currently not used.
feature_type: Describes the feature type, i.e. factor or numeric.
levels: The class levels of categorical features. This is used to check prospective datasets.
ordered: Specifies whether the
distribution: Five-number summary (numeric) or class frequency (categorical).
data_id: Internal identifier for the dataset used to derive the feature information.
run_id: Internal identifier for the specific subset of the dataset used to derive the feature information.
in_signature: Specifies whether the feature is included in the model signature.
in_novelty: Specifies whether the feature is included in the novelty detector.
removed: Specifies whether the feature was removed during pre-processing.
removed_unknown_type: Specifies whether the feature was removed during pre-processing because the type was neither factor nor numeric..
removed_missing_values: Specifies whether the feature was removed during pre-processing because it contained too many missing values.
removed_no_variance: Specifies whether the feature was removed during pre-processing because it did not contain more than 1 unique value.
removed_low_variance: Specifies whether the feature was removed during pre-processing because the variance was too low. Requires applying low_variance as a filter_method.
removed_low_robustness: Specifies whether the feature was removed during pre-processing because it lacks robustness. Requires applying robustness as a filter_method, as well as repeated measurement.
removed_low_importance: Specifies whether the feature was removed during pre-processing because it lacks relevance. Requires applying univariate_test as a filter_method.
fraction_missing: Specifies the fraction of missing values.
robustness: Specifies robustness of the feature, if measured.
univariate_importance: Specifies the univariate p-value of the feature, if measured.
transformation_parameters: Details parameters for power transformation of numeric features.
normalisation_parameters: Details parameters for (global) normalisation of numeric features.
batch_normalisation_parameters: Details parameters for batch normalisation of numeric features.
imputation_parameters: Details parameters or models for imputation of missing values.
cluster_parameters: Details parameters for forming clusters with other features.
required_features: Details features required for clustering or imputation.
familiar_version: Version of the familiar package.