Implementation:Scikit learn Scikit learn RANSACRegressor

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Machine Learning, Robust Regression
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for robust parameter estimation using the RANSAC (RANdom SAmple Consensus) algorithm provided by scikit-learn.

Description

RANSACRegressor implements the RANSAC algorithm for iteratively estimating model parameters from a subset of inliers. At each iteration, a random subset of samples is selected, a base estimator (default: LinearRegression) is fit on this subset, and the remaining samples are tested against the model to identify inliers based on a residual threshold. The process repeats, keeping the model with the largest consensus set (most inliers). RANSAC is a meta-estimator that wraps any regression estimator supporting fit, score, and predict methods.

Usage

Use RANSACRegressor when your dataset contains significant outliers that would corrupt standard regression methods. It is particularly effective when you know that a substantial fraction of the data consists of inliers that follow the model, while the outliers are arbitrary. Common applications include computer vision (line/plane fitting), sensor data processing, and any regression task with gross outliers.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/linear_model/_ransac.py

Signature

class RANSACRegressor(
    MetaEstimatorMixin,
    RegressorMixin,
    MultiOutputMixin,
    BaseEstimator,
):
    def __init__(
        self,
        estimator=None,
        *,
        min_samples=None,
        residual_threshold=None,
        is_data_valid=None,
        is_model_valid=None,
        max_trials=100,
        max_skips=np.inf,
        stop_n_inliers=np.inf,
        stop_score=np.inf,
        stop_probability=0.99,
        loss="absolute_error",
        random_state=None,
    ):

Import

from sklearn.linear_model import RANSACRegressor

I/O Contract

Inputs

Name	Type	Required	Description
estimator	object	No	Base estimator with fit/score/predict methods (default=LinearRegression)
min_samples	int or float	No	Minimum random samples for fitting; absolute if >=1, relative if <1 (default=None)
residual_threshold	float	No	Maximum residual for inlier classification; defaults to MAD of y (default=None)
is_data_valid	callable	No	Function called with random subset to validate data (default=None)
is_model_valid	callable	No	Function called with model and random subset to validate model (default=None)
max_trials	int	No	Maximum number of random sample iterations (default=100)
stop_n_inliers	int	No	Stop if at least this many inliers are found (default=inf)
stop_score	float	No	Stop if score is at least this value (default=inf)
stop_probability	float	No	Confidence probability for early stopping of trials (default=0.99)
loss	str or callable	No	Loss function: 'absolute_error', 'squared_error', or callable (default='absolute_error')
random_state	int or RandomState	No	Random seed for reproducibility

Outputs

Name	Type	Description
estimator_	object	Best fitted base estimator on the final inlier set
n_trials_	int	Number of random selection trials performed
inlier_mask_	ndarray of shape (n_samples,)	Boolean mask of inlier samples
n_skips_no_inliers_	int	Number of iterations skipped due to finding no inliers
n_skips_invalid_data_	int	Number of iterations skipped due to invalid data
n_skips_invalid_model_	int	Number of iterations skipped due to invalid model

Usage Examples

Basic Usage

from sklearn.linear_model import RANSACRegressor
from sklearn.datasets import make_regression
import numpy as np

X, y = make_regression(n_samples=200, n_features=5, noise=10, random_state=42)
# Add outliers
y[:20] = np.random.RandomState(42).uniform(-500, 500, size=20)

model = RANSACRegressor(random_state=42)
model.fit(X, y)
print("Inliers:", model.inlier_mask_.sum(), "out of", len(y))
print("Trials:", model.n_trials_)

Related Pages

• Principle:Scikit_learn_Scikit_learn_Robust_Regression