Transformer that filters out columns specified in columnsToOmit.
Transformer that changes column name.
Transformer that filters out prediction column and renames predictedLabels column to prediction column.
Some spark operation assume their input was string-indexed.
Model wrapper adding 'string indexing' behaviour.
Model wrapper adding 'string indexing' behaviour.
Concrete models (like GBTClassificationModel) must be concrete classes (leaves in hierarchy). That's why this class must be abstract.
Some spark operation assume their input was string-indexed. User-experience suffers from this requirement. We can work around it by wrapping estimation in
StringIndexerEstimatorWrapper
.StringIndexerEstimatorWrapper
plugs in StringIndexer before operation. It also makes it transparent for clients' components by reverting string indexing with labelConverter.