SparkBlender
- class sparklightautoml.automl.blend.SparkBlender[source]
Bases:
TransformerInputOutputRoles
,ABC
Basic class for blending.
Blender learns how to make blend on sequence of prediction datasets and prune pipes, that are not used in final blend.
- property input_roles
Returns dict of input roles
- property output_roles
Returns dict of output roles
- transformer(*args, **kwargs)[source]
Returns Spark MLlib Transformer. Represents a Transformer with fitted models.
- Return type:
Optional
[Transformer
]
- split_models(predictions, pipes)[source]
Split predictions by single model prediction datasets.
- Parameters:
predictions (
SparkDataset
) – Dataset with predictions.pipes (
Sequence
[SparkMLPipeline
]) – ml pipelines to be associated with the predictions
- Returns:
prediction column name
corresponding model index (in the pipe)
corresponding pipe index
- Return type:
Each tuple in the list is
- score(dataset)[source]
Score metric for blender.
- Parameters:
dataset (
SparkDataset
) – Blended predictions dataset.- Return type:
- Returns:
Metric value.