Implementation:Scikit learn Scikit learn Birch

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Clustering, Online Learning
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for performing memory-efficient online-learning clustering using the BIRCH algorithm provided by scikit-learn.

Description

Birch (Balanced Iterative Reducing and Clustering using Hierarchies) is an online-learning clustering algorithm that builds a tree data structure called a Clustering Feature (CF) tree. It incrementally processes data points, inserting them into a compact summary representation. The leaf nodes of the CF tree contain subclusters whose centroids can serve as final cluster centers or as input to another clustering algorithm such as AgglomerativeClustering. It is designed as a memory-efficient alternative to MiniBatchKMeans.

Usage

Use Birch when you have large datasets that cannot fit in memory, when you need an online (incremental) clustering algorithm, or when you want a first pass to reduce data dimensionality before applying a more expensive clustering algorithm. It is particularly effective for data that has a natural spherical cluster structure.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/cluster/_birch.py

Signature

class Birch(
    ClassNamePrefixFeaturesOutMixin, ClusterMixin, TransformerMixin, BaseEstimator
):
    def __init__(
        self,
        *,
        threshold=0.5,
        branching_factor=50,
        n_clusters=3,
        compute_labels=True,
    ):

Import

from sklearn.cluster import Birch

I/O Contract

Inputs

Name	Type	Required	Description
threshold	float	No	Maximum radius of a subcluster for merging. Lower values promote splitting. Default is 0.5.
branching_factor	int	No	Maximum number of CF subclusters in each node. Default is 50.
n_clusters	int, sklearn.cluster model, or None	No	Number of clusters after the final clustering step. Can also be a clustering model instance. Default is 3.
compute_labels	bool	No	Whether to compute labels for each fit. Default is True.

Outputs

Name	Type	Description
root_	_CFNode	Root of the CF tree.
dummy_leaf_	_CFNode	Pointer to the first leaf in the CF tree.
subcluster_centers_	ndarray	Centroids of all subclusters read from leaves.
subcluster_labels_	ndarray	Labels assigned to the subcluster centroids.
labels_	ndarray of shape (n_samples,)	Cluster labels for each sample (if compute_labels=True).

Usage Examples

Basic Usage

from sklearn.cluster import Birch
import numpy as np

X = np.array([[1, 2], [1, 4], [1, 0],
              [10, 2], [10, 4], [10, 0]])

brc = Birch(n_clusters=2)
brc.fit(X)
print(brc.labels_)
print(brc.subcluster_centers_)

Related Pages

• Principle:Scikit_learn_Scikit_learn_Clustering