Implementation:Scikit learn Scikit learn ComputeClassWeight

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

Overview

Concrete tool for estimating class weights for unbalanced datasets, provided by scikit-learn.

Description

The compute_class_weight function estimates class weights for unbalanced datasets. When set to "balanced", weights are computed as n_samples / (n_classes * np.bincount(y)), inversely proportional to class frequencies. It also supports custom weight dictionaries and optional sample weights. The "balanced" heuristic is inspired by King and Zen (2001).

Usage

Use this function to compute class weights that can be passed to classifiers (via the class_weight parameter) to handle class imbalance in classification tasks.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/utils/class_weight.py

Signature

@validate_params(...)
def compute_class_weight(class_weight, *, classes, y, sample_weight=None):

Import

from sklearn.utils.class_weight import compute_class_weight

I/O Contract

Inputs

Name	Type	Required	Description
class_weight	dict, "balanced", or None	Yes	Weight strategy: dict of class-to-weight, "balanced" for auto-computed, or None for uniform
classes	ndarray	Yes	Array of classes occurring in the data
y	array-like of shape (n_samples,)	Yes	Array of original class labels per sample
sample_weight	array-like of shape (n_samples,) or None	No	Per-sample weights, only used when class_weight='balanced' (default None)

Outputs

Name	Type	Description
class_weight_vect	ndarray of shape (n_classes,)	Array of computed class weights

Usage Examples

Basic Usage

import numpy as np
from sklearn.utils.class_weight import compute_class_weight

y = [1, 1, 1, 1, 0, 0]
classes = np.unique(y)
weights = compute_class_weight("balanced", classes=classes, y=y)
print(weights)  # [1.5, 0.75] - higher weight for minority class

Related Pages

• Principle:Scikit_learn_Scikit_learn_Feature_Encoding