PowerTransformer

class ray.data.preprocessors.PowerTransformer(columns: List[str], power: float, method: str = 'yeo-johnson', *, output_columns: List[str] | None = None)

Bases: SerializablePreprocessorBase

Apply a power transform to make your data more normally distributed.

Some models expect data to be normally distributed. By making your data more Gaussian-like, you might be able to improve your model’s performance.

This preprocessor supports the following transformations:

• Yeo-Johnson

• Box-Cox

Box-Cox requires all data to be positive.

Warning

You need to manually specify the transform’s power parameter. If you choose a bad value, the transformation might not work well.

Parameters:

• columns (List[str]) – The columns to separately transform.

• power (float) – A parameter that determines how your data is transformed. Practioners typically set power between \(-2.5\) and \(2.5\), although you may need to try different values to find one that works well.

• method (str) – A string representing which transformation to apply. Supports "yeo-johnson" and "box-cox". If you choose "box-cox", your data needs to be positive. Defaults to "yeo-johnson".

• output_columns (List[str] | None) – The names of the transformed columns. If None, the transformed columns will be the same as the input columns. If not None, the length of output_columns must match the length of columns, othwerwise an error will be raised.

PublicAPI (alpha): This API is in alpha and may change before becoming stable.

Methods

deserialize	Deserialize a preprocessor from serialized data.
fit	Fit this Preprocessor to the Dataset.
fit_status	Return the fit status of this preprocessor.
fit_transform	Fit this Preprocessor to the Dataset and then transform the Dataset.
get_input_columns	Return the names of the columns this preprocessor reads.
get_output_columns	Return the names of the columns this preprocessor writes.
get_preprocessor_class_id	Get the preprocessor class identifier for this preprocessor class.
get_version	Get the version number for this preprocessor class.
has_stats	Return whether this preprocessor holds computed statistics.
preferred_batch_format	Batch format hint for upstream producers to try yielding best block format.
serialize	Serialize this preprocessor to a string or bytes.
set_preprocessor_class_id	Set the preprocessor class identifier for this preprocessor class.
set_version	Set the version number for this preprocessor class.
transform	Transform the given dataset.
transform_batch	Transform a single batch of data.

Attributes

MAGIC_CLOUDPICKLE
SERIALIZER_FORMAT_VERSION
columns	The columns to separately transform.
method	"yeo-johnson" or "box-cox".
output_columns	The names of the transformed columns.
power	The parameter that determines how the data is transformed.
stat_computation_plan	The aggregations that fit() runs to compute statistics.