Sklearn Is Fitted¶
========================================
__sklearn_is_fitted__ as Developer API¶
The __sklearn_is_fitted__ method is a convention used in scikit-learn for
checking whether an estimator object has been fitted or not. This method is
typically implemented in custom estimator classes that are built on top of
scikit-learn’s base classes like BaseEstimator or its subclasses.
Developers should use :func:~sklearn.utils.validation.check_is_fitted
at the beginning of all methods except fit. If they need to customize or
speed-up the check, they can implement the __sklearn_is_fitted__ method as
shown below.
In this example the custom estimator showcases the usage of the
__sklearn_is_fitted__ method and the check_is_fitted utility function
as developer APIs. The __sklearn_is_fitted__ method checks fitted status
by verifying the presence of the _is_fitted attribute.
Imports for Implementing the __sklearn_is_fitted__ Developer API¶
check_is_fitted and __sklearn_is_fitted__ form the contract between scikit-learn’s infrastructure and custom estimators for guarding against calling predict or transform before fit: By default, check_is_fitted inspects the estimator for attributes ending with _ (like coef_ or classes_) that are conventionally set during fit. However, custom estimators can override this behavior by implementing __sklearn_is_fitted__ to return a boolean, enabling faster or more precise fitted-status checks – for example, checking a single _is_fitted flag instead of scanning all attributes.
This pattern is essential for developing scikit-learn-compatible estimators that integrate with pipelines, cross-validation, and meta-estimators: When a custom estimator is used inside Pipeline, GridSearchCV, or VotingClassifier, scikit-learn internally calls check_is_fitted before routing data through predict or transform. If the check fails, a NotFittedError is raised with a clear message, preventing silent bugs from unfitted models producing meaningless predictions. The BaseEstimator and ClassifierMixin base classes provide the boilerplate for get_params, set_params, and score, while the developer is responsible for implementing fit, predict, and the fitted-status protocol.
# %%
# An example custom estimator implementing a simple classifier
# ------------------------------------------------------------
# This code snippet defines a custom estimator class called `CustomEstimator`
# that extends both the `BaseEstimator` and `ClassifierMixin` classes from
# scikit-learn and showcases the usage of the `__sklearn_is_fitted__` method
# and the `check_is_fitted` utility function.
# Authors: The scikit-learn developers
# SPDX-License-Identifier: BSD-3-Clause
from sklearn.base import BaseEstimator, ClassifierMixin
from sklearn.utils.validation import check_is_fitted
class CustomEstimator(BaseEstimator, ClassifierMixin):
def __init__(self, parameter=1):
self.parameter = parameter
def fit(self, X, y):
"""
Fit the estimator to the training data.
"""
self.classes_ = sorted(set(y))
# Custom attribute to track if the estimator is fitted
self._is_fitted = True
return self
def predict(self, X):
"""
Perform Predictions
If the estimator is not fitted, then raise NotFittedError
"""
check_is_fitted(self)
# Perform prediction logic
predictions = [self.classes_[0]] * len(X)
return predictions
def score(self, X, y):
"""
Calculate Score
If the estimator is not fitted, then raise NotFittedError
"""
check_is_fitted(self)
# Perform scoring logic
return 0.5
def __sklearn_is_fitted__(self):
"""
Check fitted status and return a Boolean value.
"""
return hasattr(self, "_is_fitted") and self._is_fitted