Custom Transformers

Custom transformers are user-defined functions that transform data in an XPipeline. They are defined in a separate file and imported into a Python session.

Custom transformers inherit from the XBaseTransformer class in the xplainable package. The XBaseTransformer class provides several methods you can override to customize the transformer’s behaviour. The most important of these is the transform method, which every transformer in a pipeline will call.

You can find details about the XBaseTransformer class in the API documentation.

Example

To create a custom transformer, create a Python file in your working directory called custom_transformers.py. In this file, define a class that inherits from XBaseTransformer:

from xplainable.core.preprocessing.base import XBaseTransformer

class MyTransformer(XBaseTransformer):

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    # Optional
    def fit(self, data):
        # learn something from the data
        return data

    # Required
    def transform(self, data):
        # do something to the data
        return data

    # Optional
    def inverse_transform(self, data):
        # reverse the transformation
        return data

You can then import this class into a Python session and use it in an XPipeline:

from xplainable.preprocessing.pipeline import XPipeline
from custom_transformers import MyTransformer

pipeline = XPipeline([
    {"transformer": MyTransformer()},
])