- class sklearn.cluster.AgglomerativeClustering(n_clusters=2, *, metric='euclidean', memory=None, connectivity=None, compute_full_tree='auto', linkage='ward', distance_threshold=None, compute_distances=False)[source]#
Agglomerative Clustering.
Recursively merges pair of clusters of sample data; uses linkage distance.
Read more in the User Guide.
- Parameters:
- n_clustersint or None, default=2
The number of clusters to find. It must be
None
ifdistance_threshold
is notNone
.- metricstr or callable, default=”euclidean”
Metric used to compute the linkage. Can be “euclidean”, “l1”, “l2”,“manhattan”, “cosine”, or “precomputed”. If linkage is “ward”, only“euclidean” is accepted. If “precomputed”, a distance matrix is neededas input for the fit method.
Added in version 1.2.
Deprecated since version 1.4:
metric=None
is deprecated in 1.4 and will be removed in 1.6.Letmetric
be the default value (i.e."euclidean"
) instead.- memorystr or object with the joblib.Memory interface, default=None
Used to cache the output of the computation of the tree.By default, no caching is done. If a string is given, it is thepath to the caching directory.
- connectivityarray-like, sparse matrix, or callable, default=None
Connectivity matrix. Defines for each sample the neighboringsamples following a given structure of the data.This can be a connectivity matrix itself or a callable that transformsthe data into a connectivity matrix, such as derived from
kneighbors_graph
. Default isNone
, i.e, thehierarchical clustering algorithm is unstructured.- compute_full_tree‘auto’ or bool, default=’auto’
Stop early the construction of the tree at
n_clusters
. This isuseful to decrease computation time if the number of clusters is notsmall compared to the number of samples. This option is useful onlywhen specifying a connectivity matrix. Note also that when varying thenumber of clusters and using caching, it may be advantageous to computethe full tree. It must beTrue
ifdistance_threshold
is notNone
. By defaultcompute_full_tree
is “auto”, which is equivalenttoTrue
whendistance_threshold
is notNone
or thatn_clusters
is inferior to the maximum between 100 or0.02 * n_samples
.Otherwise, “auto” is equivalent toFalse
.- linkage{‘ward’, ‘complete’, ‘average’, ‘single’}, default=’ward’
Which linkage criterion to use. The linkage criterion determines whichdistance to use between sets of observation. The algorithm will mergethe pairs of cluster that minimize this criterion.
‘ward’ minimizes the variance of the clusters being merged.
‘average’ uses the average of the distances of each observation ofthe two sets.
‘complete’ or ‘maximum’ linkage uses the maximum distances betweenall observations of the two sets.
‘single’ uses the minimum of the distances between all observationsof the two sets.
Added in version 0.20: Added the ‘single’ option
For examples comparing different
linkage
criteria, seeComparing different hierarchical linkage methods on toy datasets.- distance_thresholdfloat, default=None
The linkage distance threshold at or above which clusters will not bemerged. If not
None
,n_clusters
must beNone
andcompute_full_tree
must beTrue
.Added in version 0.21.
- compute_distancesbool, default=False
Computes distances between clusters even if
distance_threshold
is notused. This can be used to make dendrogram visualization, but introducesa computational and memory overhead.Added in version 0.24.
For an example of dendrogram visualization, seePlot Hierarchical Clustering Dendrogram.
- Attributes:
- n_clusters_int
The number of clusters found by the algorithm. If
distance_threshold=None
, it will be equal to the givenn_clusters
.- labels_ndarray of shape (n_samples)
Cluster labels for each point.
- n_leaves_int
Number of leaves in the hierarchical tree.
- n_connected_components_int
The estimated number of connected components in the graph.
Added in version 0.21:
n_connected_components_
was added to replacen_components_
.- n_features_in_int
Number of features seen during fit.
Added in version 0.24.
- feature_names_in_ndarray of shape (
n_features_in_
,) Names of features seen during fit. Defined only when
X
has feature names that are all strings.Added in version 1.0.
- children_array-like of shape (n_samples-1, 2)
The children of each non-leaf node. Values less than
n_samples
correspond to leaves of the tree which are the original samples.A nodei
greater than or equal ton_samples
is a non-leafnode and has childrenchildren_[i - n_samples]
. Alternativelyat the i-th iteration, children[i][0] and children[i][1]are merged to form noden_samples + i
.- distances_array-like of shape (n_nodes-1,)
Distances between nodes in the corresponding place in
children_
.Only computed ifdistance_threshold
is used orcompute_distances
is set toTrue
.
See also
- FeatureAgglomeration
Agglomerative clustering but for features instead of samples.
- ward_tree
Hierarchical clustering with ward linkage.
Examples
>>> from sklearn.cluster import AgglomerativeClustering>>> import numpy as np>>> X = np.array([[1, 2], [1, 4], [1, 0],... [4, 2], [4, 4], [4, 0]])>>> clustering = AgglomerativeClustering().fit(X)>>> clusteringAgglomerativeClustering()>>> clustering.labels_array([1, 1, 1, 0, 0, 0])
- fit(X, y=None)[source]#
Fit the hierarchical clustering from features, or distance matrix.
- Parameters:
- Xarray-like, shape (n_samples, n_features) or (n_samples, n_samples)
Training instances to cluster, or distances between instances if
metric='precomputed'
.- yIgnored
Not used, present here for API consistency by convention.
- Returns:
- selfobject
Returns the fitted instance.
- fit_predict(X, y=None)[source]#
Fit and return the result of each sample’s clustering assignment.
In addition to fitting, this method also return the result of theclustering assignment for each sample in the training set.
- Parameters:
- Xarray-like of shape (n_samples, n_features) or (n_samples, n_samples)
Training instances to cluster, or distances between instances if
affinity='precomputed'
.- yIgnored
Not used, present here for API consistency by convention.
- Returns:
- labelsndarray of shape (n_samples,)
Cluster labels.
- get_metadata_routing()[source]#
Get metadata routing of this object.
Please check User Guide on how the routingmechanism works.
- Returns:
- routingMetadataRequest
A MetadataRequest encapsulatingrouting information.
- get_params(deep=True)[source]#
Get parameters for this estimator.
- Parameters:
- deepbool, default=True
If True, will return the parameters for this estimator andcontained subobjects that are estimators.
- Returns:
- paramsdict
Parameter names mapped to their values.
- set_params(**params)[source]#
Set the parameters of this estimator.
The method works on simple estimators as well as on nested objects(such as Pipeline). The latter haveparameters of the form
<component>__<parameter>
so that it’spossible to update each component of a nested object.- Parameters:
- **paramsdict
Estimator parameters.
- Returns:
- selfestimator instance
Estimator instance.
Gallery examples#
A demo of structured Ward hierarchical clustering on an image of coins
A demo of structured Ward hierarchical clustering on an image of coins
Agglomerative clustering with and without structure
Agglomerative clustering with and without structure
Agglomerative clustering with different metrics
Agglomerative clustering with different metrics
Comparing different clustering algorithms on toy datasets
Comparing different clustering algorithms on toy datasets
Comparing different hierarchical linkage methods on toy datasets
Comparing different hierarchical linkage methods on toy datasets
Hierarchical clustering: structured vs unstructured ward
Hierarchical clustering: structured vs unstructured ward
Inductive Clustering
Inductive Clustering
Plot Hierarchical Clustering Dendrogram
Plot Hierarchical Clustering Dendrogram
Various Agglomerative Clustering on a 2D embedding of digits
Various Agglomerative Clustering on a 2D embedding of digits