🕸️ Community Detection in Graphs

A Python project for analyzing network structures and detecting communities using various algorithms such as Louvain, Leiden, Infomap, and others.
The project supports both directed and undirected graphs, computes network statistics, detects communities, and exports comparative evaluation results.

---

📁 Project Structure

community-detection/
├── cd_benchmark/       # Core modules for analysis, detection, statistics, visualization
├── run/                # Executable scripts (entry points)
├── data/               # Input graph datasets
├── examples/           # Example usage scripts and visualizations
├── requirements.txt    # Python dependencies
├── pyproject.toml      # Package configuration
├── README.md           # Project documentation

---

Getting Started

Installation

Option 1: Install from PyPI (coming soon)

pip install cd-benchmark

Option 2: Install from source

git clone https://github.com/rpritr/network-community-detection-benchmark.git
cd community-detection
pip install -e .

Option 3: Development setup

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Quick Start

After installation, you can use the package in your Python code:

from cd_benchmark.analysis import CommunityAnalysis
import networkx as nx

# Create or load your graph
G = nx.karate_club_graph()

# Run community detection
ca = CommunityAnalysis(graph=G)
results = ca.run(algorithms=["Louvain", "Leiden", "Infomap"])
print(results)

Run Examples

python -m run.run_statistics

Or run the full benchmarking with configurable algorithms:

python -m run.run_analysis

---

🔧 Features

• Load graphs from edge list text files (directed or undirected)
• Compute network statistics:
  • Node and edge count
  • Largest WCC/SCC components
  • Clustering coefficient
  • Graph density, diameter, and radius
• Community detection algorithms:
  • Louvain
  • Leiden
  • Infomap
  • Walktrap
  • Girvan-Newman
  • Greedy modularity
  • Label Propagation
  • Fast Label Propagation
• Graph visualization (with Matplotlib / NetworkX)
• Export results to CSV for analysis and comparison

---

Community Analysis

You can instantiate and run custom analyses as follows:

from cd_benchmark.analysis import CommunityAnalysis

ca = CommunityAnalysis(graph=G)  # or file="data/graph.txt"
df = ca.run(algorithms=["Louvain", "Infomap"])

Supported algorithms:

["Louvain", "Leiden", "Infomap", "Girvan Newman", "Greedy Modularity", "Walktrap", "Label Propagation", "Fast Label Propagation"]

---

Community Benchmark

You can run a full benchmark on 100 iterations as follows:

from cd_benchmark.benchmark import CommunityBenchmark

cb = CommunityBenchmark(graph=G)  # or file="data/graph.txt"
df = cb.run()
cb.summarize()
cb.plot_all()  # plot and export analysis graphs

Sample Datasets

Example graph format (cit-Patents.txt):

Node1 Node2
12    14
15    17
...

Each line represents a directed edge in the network.

---

Dependencies

Install all required packages from requirements.txt:

Note: Some community detection methods in cdlib may require additional system-level libraries such as graph-tool or infomap.

---

Author

Robi Pritržnik (2025)
🔗 pritrznik.si
📧 Contact: robi@pritrznik.si

---

License

This project is intended for research and educational purposes only.
Feel free to fork and extend under appropriate attribution.

---

Dataset Source

Sample networks in this project are based on the SNAP collection:

Leskovec, J., & Krevl, A. (2014). SNAP Datasets: Stanford Large Network Dataset Collection. Retrieved from http://snap.stanford.edu/data

Notes on AI Assistance Disclosure

Parts of this project ( code refactoring and documentation) were developed with the assistance of ChatGPT, based on the GPT-4o model (OpenAI, August 2025).