Facebook Ego Network Analysis for Political Outreach Optimization

A data-driven approach to optimizing political outreach campaigns using network science and community detection algorithms.

📌 Overview

For Employers: See EMPLOYER_HIGHLIGHTS.md for a quick summary of skills demonstrated and business impact.

This project applies advanced network analysis techniques to identify influential users and communities within social networks, enabling more efficient and cost-effective political outreach strategies. By leveraging the Facebook ego network dataset from Stanford SNAP, this analysis demonstrates how targeted approaches can achieve 92.2% outreach coverage compared to just 25.2% with random targeting.

Key Applications

• Political Campaigns: Optimize advertising spend by targeting high-influence individuals
• Social Media Marketing: Identify key influencers and community structures
• Information Diffusion: Model and predict how information spreads through networks
• Community Analysis: Understand social group dynamics and relationships

🔑 Key Features

Network Analysis

• Structural Metrics: Clustering coefficient (C = 0.51), average path length (L = 3.75)
• Degree Distribution: Identifies hub nodes and network topology
• Small-World Properties: Analyzes network efficiency and connectivity

Community Detection

• Louvain Algorithm: Detects 13 algorithmic communities with varying densities
• Comparison with Ground Truth: Evaluates against user-defined social circles
• Adjusted Rand Index (ARI): Measures alignment between detection methods (ARI = 0.221)

Centrality Analysis

• Degree Centrality: Identifies highly connected influencers
• Betweenness Centrality: Discovers bridge nodes connecting different communities
• Eigenvector Centrality: Finds nodes connected to other important nodes

Influence Propagation

• Threshold Model: Simulates information spread with configurable activation thresholds
• Targeted vs. Random Seeding: Compares strategic vs. random user targeting
• ROI Analysis: Demonstrates significant efficiency gains with data-driven approaches

🚀 Quick Start

New to this project? See QUICKSTART.md for a 5-minute setup guide.

Prerequisites

• Python 3.8 or higher
• pip package manager

Installation

1. Clone this repository:

git clone https://github.com/aymanmomin/Facebook-Ego-Network-Analysis-for-Political-Outreach-Optimization.git
cd Facebook-Ego-Network-Analysis-for-Political-Outreach-Optimization

2. Install required dependencies:

pip install -r requirements.txt

3. The dataset is included in the data/ directory (Stanford SNAP Ego-Facebook dataset)

Usage

Open and run the Jupyter notebook:

# Main analysis notebook
jupyter notebook cpsc-572-facebook-ego-network-analysis.ipynb

Or use JupyterLab:

jupyter lab

The notebook is fully self-contained and includes:

• Automatic package installation
• Data loading and preprocessing
• All analysis steps with detailed explanations
• Visualization generation

📊 Results & Insights

Community Structure

• 13 Communities Detected: Varying from tight-knit groups (density = 1.0) to loose networks (density = 0.154)
• High-Density Communities: Enable rapid information spread within groups
• Low-Density Communities: Require bridge nodes for effective reach

Influence Campaigns

• Targeted Approach: 92.2% network coverage with strategic seeding
• Random Approach: Only 25.2% coverage with same budget
• Key Finding: Degree centrality is the most effective metric for practical targeting

Bridge Nodes

Critical connectors identified (nodes 277, 175, 19, 23, 25) that link otherwise disconnected communities, essential for cross-community influence.

📁 Project Structure

├── notebooks/
│   └── facebook-ego-network-analysis.ipynb  # Main analysis notebook
├── data/                          # Facebook ego network dataset
│   ├── facebook_combined.txt      # Combined edge list
│   ├── *.edges                    # Individual ego networks
│   ├── *.circles                  # User-defined social circles
│   ├── *.feat                     # Node features
│   └── network_communities.html   # Interactive visualization
├── visualizations/                # Analysis output images
│   ├── day1_results.png           # Basic statistics
│   ├── day2_results.png           # Community detection
│   ├── day3_results.png           # Influence simulation
│   ├── part1_results.png
│   └── part2_results.png
├── docs/                          # Documentation
│   ├── README.md                  # Documentation index
│   └── technical-report.md        # Detailed analysis report
├── requirements.txt               # Python dependencies
├── CONTRIBUTING.md                # Contribution guidelines
├── REORGANIZATION_GUIDE.md        # File organization instructions
├── README.md                      # This file
├── LICENSE                        # MIT License
└── .gitignore                     # Git ignore patterns

Note: See REORGANIZATION_GUIDE.md for instructions on organizing files into the recommended structure.

🛠️ Technical Stack

• NetworkX: Graph manipulation and analysis
• Python-Louvain: Community detection
• Pandas: Data manipulation
• NumPy: Numerical computations
• Matplotlib: Static visualizations
• PyVis: Interactive network visualizations
• Scikit-learn: Similarity metrics (ARI)
• tqdm: Progress tracking

📈 Methodology

1. Data Preprocessing: Load ego network, filter isolates, construct graph
2. Structural Analysis: Calculate network metrics and compare with null models
3. Community Detection: Apply Louvain algorithm and compare with ground truth
4. Centrality Computation: Rank nodes by influence potential
5. Threshold Simulation: Model information diffusion with activation thresholds
6. Strategy Comparison: Evaluate targeted vs. random seeding approaches

🎯 Use Cases

This framework can be adapted for:

• Marketing Campaigns: Identify brand ambassadors and influencers
• Public Health: Optimize health intervention programs
• Corporate Communications: Internal information dissemination strategies
• Academic Research: Social network studies and graph theory
• Product Launches: Viral marketing and early adopter identification

📚 Dataset

Stanford SNAP Ego-Facebook Network

• Source: Stanford Network Analysis Project
• Description: Anonymized Facebook friendship networks from survey participants
• Size: 4,039 nodes, 88,234 edges
• Features: Anonymized profile attributes, user-defined circles
• Citation: J. McAuley and J. Leskovec. Learning to Discover Social Circles in Ego Networks. NIPS, 2012.

🤝 Contributing

Contributions are welcome! Feel free to:

• Open issues for bugs or feature requests
• Submit pull requests with improvements
• Share your use cases and adaptations

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

👤 Author

Ayman Momin

• GitHub: @aymanmomin
• LinkedIn: Connect with me

📖 Additional Resources

• RESUME_EXAMPLES.md - Ready-to-use resume bullets and LinkedIn posts
• EMPLOYER_HIGHLIGHTS.md - Executive summary for recruiters
• QUICKSTART.md - 5-minute setup guide
• Report.md - Detailed technical analysis

🙏 Acknowledgments

• Stanford Network Analysis Project (SNAP) for providing the dataset
• NetworkX development team for the excellent graph analysis library
• Python open-source community

---

⭐ If you find this project useful, please consider giving it a star!