ccluster 0.1.1

A package for clustering with cluster size constraints.

ccluster: a package for clustering with size constraints

ccluster is a library for performing clustering with exhaustive or partial cluster size constraints. It provides two constrained clustering algorithms: a constrained k-means suitable for euclidean data, and a constrained graph clustering algorithm based on spectral clustering.

Quick-Start

Here is an example of using constrained $k$-means with exhaustive constraints i.e. giving a size constraint for each cluster:

>>> from ccluster.size import ConstrainedKMeans
>>> import numpy as np

>>> X = np.array([[1, 2], [1, 4], [1, 0],
...               [10, 2], [10, 4], [10, 0]])
>>> kmeans = ConstrainedKMeans(
...     n_clusters=2,
...     cluster_size=[2, 4],
...     random_state=0,
...     n_init=10).fit(X)
>>> kmeans.labels_
# array([1, 1, 1, 0, 0, 1])  
>>> kmeans.predict([[0, 0], [12, 3]], [1, 1])
# array([1, 0], dtype=int32)  
>>> kmeans.cluster_centers_
# array([[10. , 3. ],  
#        [ 3.25, 1.5 ]])  

It is also possible to specify partial constraints i.e. constraints on a subset of the clusters and leave the remaining ones free. Here is an example using constrained spectral clustering:

>>> from ccluster.size import ConstrainedSpectralClustering
>>> import numpy as np

>>> X = np.array([[1, 1], [2, 1], [1, 0],
...               [4, 7], [3, 5], [3, 6],
...               [9, 6], [5, 4], [2, 1]])

>>> spectral = ConstrainedSpectralClustering(
...     n_clusters=4, 
...     cluster_sizes=[2, 2],
...     random_state=0).fit(X)

>>> spectral.labels_
# array([2, 3, 2, 0, 3, 0, 1, 1, 3])