GeezerVison

NetworKit

High-performance tools for large-scale network analysis
Explore the docs »

Try Demo · Report Bug · Request Feature

NetworKit is an open-source toolkit for high-performance network analysis, designed to handle large networks ranging from thousands to billions of edges. Built with efficiency and scalability at its core, NetworKit implements parallel graph algorithms that leverage multicore architectures to compute standard measures of network analysis.

As both a production tool and a research testbed for algorithm engineering, NetworKit includes novel algorithms from recent publications alongside battle-tested implementations. The toolkit is available as a Python module with high-performance C++ algorithms exposed through Cython, combining Python's interactivity and rich ecosystem with C++'s computational efficiency.

(back to top)

• Scalable: Analyze networks with billions of edges
• Fast: Parallel algorithms utilizing multicore architectures
• Comprehensive: Wide range of network analysis algorithms
• Interactive: Python interface with Jupyter notebook support
• Flexible: Available as Python module or standalone C++ library
• Research-ready: Includes state-of-the-art algorithms from recent publications

(back to top)

For most users, NetworKit can be installed directly via package managers with no additional requirements other than Python 3.10+.

If you only need the C++ core without Python bindings:

More platform-specific installation instructions can be found in our getting started guide.

If you are interested in the most recent build of NetworKit, you can use the nightly channel networkit-nightly on anaconda.org. Published packages are based on pushes to the master-branch, and the channel keeps the five most recent ones.

(back to top)

Here's a quick example showing how to generate a random hyperbolic graph with 100k nodes and detect communities:

from networkit.generators import HyperbolicGenerator
from networkit.community import detectCommunities

# Generate a random hyperbolic graph
g = (
    HyperbolicGenerator(1e5)
    .generate()
)

# Detect communities
detectCommunities(g, inspect=True)

Output:

PLM(balanced,pc,turbo) detected communities in 0.14577102661132812 [s]
solution properties:
-------------------  -----------
# communities        4536
min community size      1
max community size   2790
avg. community size    22.0459
modularity              0.987243
-------------------  -----------

Compute PageRank to rank nodes by importance:

from networkit.centrality import PageRank

pr = (
    PageRank(g)
    .run()
)
top_nodes = pr.ranking()[:10]

Analyze graph structure with connected components:

from networkit.components import ConnectedComponents

cc = (
    ConnectedComponents(g)
    .run()
)
print(f"Components: {cc.numberOfComponents()}")
print(f"Largest: {max(cc.getComponentSizes().values())}")

For comprehensive examples and tutorials, explore our interactive notebooks, especially the NetworKit User Guide. You can try NetworKit directly in your browser using our Binder instance.

(back to top)

Building from source requires:

• C++ Compiler: g++ (>= 12.0), clang++ (>= 14.0), or MSVC (>= 14.30)
• OpenMP: For parallelism (usually included with compiler)
• Python: 3.10 or higher with development libraries
  • Debian/Ubuntu: apt-get install python3-dev
  • RHEL/CentOS: dnf install python3-devel
  • Windows: Official installer
• CMake: Version 3.21 or higher
• Build System: Make or Ninja

git clone https://github.com/networkit/networkit networkit
cd networkit
pip install cython numpy setuptools wheel
python setup.py build_ext [-jX]
pip install -e .

The -jX option specifies the number of threads for compilation (e.g., -j4 for 4 threads). If omitted, it uses all available CPU cores.

mkdir build && cd build
cmake ..
make -jX
sudo make install

After installation, include NetworKit headers:

#include <networkit/graph/Graph.hpp>

Compile your project:

g++ my_file.cpp -lnetworkit

To build and run tests:

cmake -DNETWORKIT_BUILD_TESTS=ON ..
make
./networkit_tests --gtest_filter=CentralityGTest.testBetweennessCentrality

For debugging with address/leak sanitizers:

cmake -DNETWORKIT_WITH_SANITIZERS=leak ..

CMakePresets.json defines the configurations that CI builds (debug, leak, release, release-native and clang-tidy), plus debug-asan for local development. Each one builds in build/<preset>:

cmake --preset debug-asan
cmake --build --preset debug-asan
ctest --preset debug-asan

The clang-tidy preset needs LLVM's Clang and clang-tidy. On macOS, add -DCMAKE_CXX_COMPILER=$(brew --prefix llvm)/bin/clang++ -DNETWORKIT_CLANG_TIDY_EXECUTABLE=$(brew --prefix llvm)/bin/clang-tidy, because clang++ is Apple's compiler and Homebrew does not put LLVM's tools on the PATH. Personal presets go into the untracked CMakeUserPresets.json.

(back to top)

The complete documentation is available online at networkit.github.io.

(back to top)

We welcome contributions to NetworKit! Whether you're fixing bugs, adding features, or improving documentation, your help makes NetworKit better for everyone.

1. Check our development guide for instructions
2. Browse open issues or open a new one
3. Fork the repository and create your feature branch
4. Submit a pull request

For support, join our mailing list.

(back to top)

Distributed under the MIT License. We ask that you cite us if you use NetworKit in your research (see our technical report and publications page).

(back to top)

• Issues: Check our issues section for existing discussions or open a new issue
• Mailing List: Subscribe here to stay updated

(back to top)

NetworKit has been used in numerous research projects. Visit our publications page for a complete list of papers about NetworKit, algorithms implemented in NetworKit, and research using NetworKit.

(back to top)

NetworKit is developed by a dedicated team of researchers and contributors. View the full list of contributors on our credits page.

(back to top)