Implementation:Scikit learn Scikit learn TransformedTargetRegressor

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Regression, Data Preprocessing
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for regressing on a transformed target variable provided by scikit-learn.

Description

TransformedTargetRegressor is a meta-estimator that applies a non-linear transformation to the target variable y in regression problems. During fit, it transforms the target (using a transformer object or a func/inverse_func pair) and fits the regressor on the transformed target. During predict, it applies the inverse transformation to the regressor's predictions. This is useful when the target distribution benefits from transformation, such as applying logarithmic transformation to skewed targets.

Usage

Use TransformedTargetRegressor when the target variable in a regression problem would benefit from transformation before fitting, such as log-transforming skewed targets or applying quantile normalization. It ensures the transformation is consistently applied during training and the inverse is correctly applied to predictions, avoiding manual transformation errors.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/compose/_target.py

Signature

class TransformedTargetRegressor(RegressorMixin, BaseEstimator):
    def __init__(
        self,
        regressor=None,
        *,
        transformer=None,
        func=None,
        inverse_func=None,
        check_inverse=True,
    ):

Import

from sklearn.compose import TransformedTargetRegressor

I/O Contract

Inputs

Name	Type	Required	Description
regressor	estimator object	No	Regressor object (e.g., derived from RegressorMixin). If None, LinearRegression is used. Default is None.
transformer	estimator object	No	Transformer object (e.g., derived from TransformerMixin). Cannot be set at the same time as func/inverse_func. Default is None.
func	callable	No	Function to apply to y before fitting. Cannot be set with transformer. Must return a 2D array. Default is None.
inverse_func	callable	No	Function to apply to predictions to return to original target space. Required if func is set. Default is None.
check_inverse	bool	No	Whether to verify that func followed by inverse_func returns the original values. Default is True.

Outputs

Name	Type	Description
y_pred	ndarray of shape (n_samples,)	Predicted target values, inverse-transformed to the original target space.
regressor_	estimator instance	The fitted regressor trained on the transformed target.
transformer_	estimator instance	The fitted transformer (either the provided transformer or a FunctionTransformer wrapping func/inverse_func).

Usage Examples

Basic Usage

from sklearn.compose import TransformedTargetRegressor
from sklearn.linear_model import LinearRegression
from sklearn.datasets import make_regression
import numpy as np

X, y = make_regression(n_samples=100, n_features=5, noise=10, random_state=42)

# Using func/inverse_func pair
regr = TransformedTargetRegressor(
    regressor=LinearRegression(),
    func=np.log1p,
    inverse_func=np.expm1,
)
# Shift target to be positive for log transform
y_positive = y - y.min() + 1
regr.fit(X, y_positive)
y_pred = regr.predict(X)
print(f"Predictions shape: {y_pred.shape}")

# Using a transformer object
from sklearn.preprocessing import QuantileTransformer
regr2 = TransformedTargetRegressor(
    regressor=LinearRegression(),
    transformer=QuantileTransformer(output_distribution="normal"),
)
regr2.fit(X, y)
y_pred2 = regr2.predict(X)
print(f"Predictions shape: {y_pred2.shape}")

Related Pages

• Principle:Scikit_learn_Scikit_learn_Linear_Regression