Research comparing overlapping communities

This repository is used to compare algorithms that can detect overlapping communities. It contains the data used and the algorithms that are compared. The code that controls the algorithms can be found in the PyAlgorithms folder. This is a Python project that calls the external buildfiles using subprocesses.

Note: This entire repository depends on linux. I have tried to make it work for Windows and Docker but without any luck.

General dependencies:

• Python 3.9 + pip
• cmake
• java
• gcc

After that run sh build_all.sh and the building steps listed below will be done automatically.

Generating synthetic networks

LFR-Benchmark can create synthetic networks given some parameters.

Building (Automatically done in shell script)

This project needs to get built so the Python code can access it.

• Install cmake on your system
• cd into src_refactor_cpp
• mkdir build
• cd build
• cmake ..

Verify build

In the README.md file of LFR-benchmark project examples can be found.

Detecting overlapping communities

There are several algorithms implemented that can run to produce a community structure of the given network with (weighted) edges.

Demon

This is a python library, listed in requirements.txt of the python project. It runs an algorithm that builds bottom to top by combining small networks.

O-HAMUHI

Is situated in a python file in directory external of the python project. The source code has not been changed.

OSLOM

It is a separate directory that contains the code. It is a c++ project that needs to get built seperately. The instructions are listed below. Oslom 2 is chosen since it is a faster version than the original Oslom, and it does give similar results.

Building OSLOM (Automatically done in shell script)

Directory: Algo_Oslom2

run sh ./compile_all.sh and this produces all the files necessary.

Again, in the readme of the project are examples that can be ran to verify success of the build.

Verifying quality of the results

NMI score

This is a c++ package that needs to get compiled.

Build instructions (Automatically done in shell script)

cd ./Measure_NMI make

To make sure it works you can run an example listed in the ReadMe.txt

Omega-Index

Package form PyPi to compute the omega index. Same API as: https://github.com/isaranto/omega_index

Running Python project

After also following the instructions listed in PyAlgorithms/Readme.md it should now be possible to run the python code. I myself used PyCharm to run the main.py file. If you run into any issues please contact us so you can also run it.