Build Status Linux MacOS Build Status Win Coverage Status PyPi Download Counter JOSS


pyclustering is a Python, C++ data mining library (clustering algorithm, oscillatory networks, neural networks). The library provides Python and C++ implementations (C++ pyclustering library) of each algorithm or model. C++ pyclustering library is a part of pyclustering and supported for Linux, Windows and MacOS operating systems.


License: The 3-Clause BSD License




PyClustering Wiki:


Required packages: scipy, matplotlib, numpy, Pillow

Python version: >=3.6 (32-bit, 64-bit)

C++ version: >= 14 (32-bit, 64-bit)


Each algorithm is implemented using Python and C/C++ language, if your platform is not supported then Python implementation is used, otherwise C/C++. Implementation can be chosen by ccore flag (by default it is always 'True' and it means that C/C++ is used), for example:

# As by default - C/C++ part of the library is used
xmeans_instance_1 = xmeans(data_points, start_centers, 20, ccore=True);

# The same - C/C++ part of the library is used by default
xmeans_instance_2 = xmeans(data_points, start_centers, 20);

# Switch off core - Python is used
xmeans_instance_3 = xmeans(data_points, start_centers, 20, ccore=False);


Installation using pip3 tool:

$ pip3 install pyclustering

Manual installation from official repository using Makefile:

# get sources of the pyclustering library, for example, from repository
$ mkdir pyclustering
$ cd pyclustering/
$ git clone .

# compile CCORE library (core of the pyclustering library).
$ cd ccore/
$ make ccore_64bit      # build for 64-bit OS

# $ make ccore_32bit    # build for 32-bit OS

# return to parent folder of the pyclustering library
$ cd ../

# install pyclustering library
$ python3 install

# optionally - test the library
$ python3 test

Manual installation using CMake:

# get sources of the pyclustering library, for example, from repository
$ mkdir pyclustering
$ cd pyclustering/
$ git clone .

# generate build files.
$ mkdir build
$ cmake ..

# build pyclustering-shared target depending on what was generated (Makefile or MSVC solution)
# if Makefile has been generated then
$ make pyclustering-shared

# return to parent folder of the pyclustering library
$ cd ../

# install pyclustering library
$ python3 install

# optionally - test the library
$ python3 test

Manual installation using Microsoft Visual Studio solution:

  1. Clone repository from:
  2. Open folder pyclustering/ccore
  3. Open Visual Studio project ccore.sln
  4. Select solution platform: x86 or x64
  5. Build pyclustering-shared project.
  6. Add pyclustering folder to python path or install it using
# install pyclustering library
$ python3 install

# optionally - test the library
$ python3 test

Proposals, Questions, Bugs

In case of any questions, proposals or bugs related to the pyclustering please contact to or create an issue here.

PyClustering Status

Branch master 0.10.1.rel
Build (Linux, MacOS) Build Status Linux MacOS Build Status Linux MacOS Build Status Linux 0.10.1.rel
Build (Win) Build Status Win Build Status Win Build Status Win 0.10.1.rel
Code Coverage Coverage Status Coverage Status Coverage Status 0.10.1.rel

Cite the Library

If you are using pyclustering library in a scientific paper, please, cite the library:

Novikov, A., 2019. PyClustering: Data Mining Library. Journal of Open Source Software, 4(36), p.1230. Available at:

BibTeX entry:

    doi         = {10.21105/joss.01230},
    url         = {},
    year        = 2019,
    month       = {apr},
    publisher   = {The Open Journal},
    volume      = {4},
    number      = {36},
    pages       = {1230},
    author      = {Andrei Novikov},
    title       = {{PyClustering}: Data Mining Library},
    journal     = {Journal of Open Source Software}

Brief Overview of the Library Content

Clustering algorithms and methods (module pyclustering.cluster):

Algorithm Python C++
Fuzzy C-Means
GA (Genetic Algorithm)

Oscillatory networks and neural networks (module pyclustering.nnet):

Model Python C++
CNN (Chaotic Neural Network)  
fSync (Oscillatory network based on Landau-Stuart equation and Kuramoto model)  
HHN (Oscillatory network based on Hodgkin-Huxley model)
Hysteresis Oscillatory Network  
LEGION (Local Excitatory Global Inhibitory Oscillatory Network)
PCNN (Pulse-Coupled Neural Network)
SOM (Self-Organized Map)
Sync (Oscillatory network based on Kuramoto model)
SyncPR (Oscillatory network for pattern recognition)
SyncSegm (Oscillatory network for image segmentation)

Graph Coloring Algorithms (module pyclustering.gcolor):

Algorithm Python C++

Containers (module pyclustering.container):

Algorithm Python C++
KD Tree
CF Tree  

Examples in the Library

The library contains examples for each algorithm and oscillatory network model:

Clustering examples: pyclustering/cluster/examples

Graph coloring examples: pyclustering/gcolor/examples

Oscillatory network examples: pyclustering/nnet/examples

Where are examples?

Code Examples

Data clustering by CURE algorithm

from pyclustering.cluster import cluster_visualizer;
from pyclustering.cluster.cure import cure;
from pyclustering.utils import read_sample;
from pyclustering.samples.definitions import FCPS_SAMPLES;

# Input data in following format [ [0.1, 0.5], [0.3, 0.1], ... ].
input_data = read_sample(FCPS_SAMPLES.SAMPLE_LSUN);

# Allocate three clusters.
cure_instance = cure(input_data, 3);
clusters = cure_instance.get_clusters();

# Visualize allocated clusters.
visualizer = cluster_visualizer();
visualizer.append_clusters(clusters, input_data);;

Data clustering by K-Means algorithm

from pyclustering.cluster.kmeans import kmeans, kmeans_visualizer
from pyclustering.cluster.center_initializer import kmeans_plusplus_initializer
from pyclustering.samples.definitions import FCPS_SAMPLES
from pyclustering.utils import read_sample

# Load list of points for cluster analysis.

# Prepare initial centers using K-Means++ method.
initial_centers = kmeans_plusplus_initializer(sample, 2).initialize()

# Create instance of K-Means algorithm with prepared centers.
kmeans_instance = kmeans(sample, initial_centers)

# Run cluster analysis and obtain results.
clusters = kmeans_instance.get_clusters()
final_centers = kmeans_instance.get_centers()

# Visualize obtained results
kmeans_visualizer.show_clusters(sample, clusters, final_centers)

Data clustering by OPTICS algorithm

from pyclustering.cluster import cluster_visualizer
from pyclustering.cluster.optics import optics, ordering_analyser, ordering_visualizer
from pyclustering.samples.definitions import FCPS_SAMPLES
from pyclustering.utils import read_sample

# Read sample for clustering from some file
sample = read_sample(FCPS_SAMPLES.SAMPLE_LSUN)

# Run cluster analysis where connectivity radius is bigger than real
radius = 2.0
neighbors = 3
amount_of_clusters = 3
optics_instance = optics(sample, radius, neighbors, amount_of_clusters)

# Performs cluster analysis

# Obtain results of clustering
clusters = optics_instance.get_clusters()
noise = optics_instance.get_noise()
ordering = optics_instance.get_ordering()

# Visualize ordering diagram
analyser = ordering_analyser(ordering)
ordering_visualizer.show_ordering_diagram(analyser, amount_of_clusters)

# Visualize clustering results
visualizer = cluster_visualizer()
visualizer.append_clusters(clusters, sample)

Simulation of oscillatory network PCNN

from pyclustering.nnet.pcnn import pcnn_network, pcnn_visualizer

# Create Pulse-Coupled neural network with 10 oscillators.
net = pcnn_network(10)

# Perform simulation during 100 steps using binary external stimulus.
dynamic = net.simulate(50, [1, 1, 1, 0, 0, 0, 0, 1, 1, 1])

# Allocate synchronous ensembles from the output dynamic.
ensembles = dynamic.allocate_sync_ensembles()

# Show output dynamic.
pcnn_visualizer.show_output_dynamic(dynamic, ensembles)

Simulation of chaotic neural network CNN

from pyclustering.cluster import cluster_visualizer
from pyclustering.samples.definitions import SIMPLE_SAMPLES
from pyclustering.utils import read_sample
from pyclustering.nnet.cnn import cnn_network, cnn_visualizer

# Load stimulus from file.
stimulus = read_sample(SIMPLE_SAMPLES.SAMPLE_SIMPLE3)

# Create chaotic neural network, amount of neurons should be equal to amount of stimulus.
network_instance = cnn_network(len(stimulus))

# Perform simulation during 100 steps.
steps = 100
output_dynamic = network_instance.simulate(steps, stimulus)

# Display output dynamic of the network.

# Display dynamic matrix and observation matrix to show clustering phenomenon.

# Visualize clustering results.
clusters = output_dynamic.allocate_sync_ensembles(10)
visualizer = cluster_visualizer()
visualizer.append_clusters(clusters, stimulus)


Cluster allocation on FCPS dataset collection by DBSCAN:

Clustering by DBSCAN

Cluster allocation by OPTICS using cluster-ordering diagram:

Clustering by OPTICS

Partial synchronization (clustering) in Sync oscillatory network:

Partial synchronization in Sync oscillatory network

Cluster visualization by SOM (Self-Organized Feature Map)

Cluster visualization by SOM