Community-Aware Portfolio Optimization in Evolving Financial Networks

This repository presents a dynamic, network-based approach to portfolio construction that leverages community detection on financial correlation networks . By identifying latent structures in asset co-movements, the strategy aims to improve diversification, reduce turnover, and enhance interpretability in portfolio design.

_{Minimum Spanning Tree of SP500 stocks.}

Table of Contents:

• Project Summary
• Project Structure
• Installation
• Notebooks

Project Summary

• Graph-based modeling of stock correlations using rolling windows
• Minimum Spanning Tree (MST) filtering to denoise correlation matrices
• Louvain community detection to uncover clusters of co-moving assets
• Event-driven rebalancing triggered by structural shifts (births, deaths, persistence changes)
• Portfolio construction using centrality-weighted representatives and shrinkage-enhanced mean–variance optimization
• Evaluation based on cumulative returns, Sharpe ratios, drawdowns, and structural metrics (entropy, persistence)

Project Structure

├── data/                   # Raw and processed financial data
├── results/                  # Figures and network plots
├── utils/                    # Source code for network construction, analysis, optimization
│   ├── data.py
│   ├── community_tools.py
│   ├── tracker.py
│   └── utils.py
├── main.py
├── community_analysis.ipynb    # Analyzes community evolution and structure
├── results_analysis.ipynb      # Visualizes and evaluates portfolio performance
├── config.yml
├── requirements.txt        # Python dependencies
└── README.md   

Installation

git clone https://github.com/SoniaBorsi/Financial-Network-Analysis.git
cd Financial-Network-Analysis
pip install -r requirements.txt

Run the main script:

python3 main.py

NOTE: All parameters for data preprocessing, community detection, and portfolio construction are set in a YAML configuration file (config.yml). This makes the pipeline easily adjustable without modifying the source code

Notebooks

• community_analysis.ipynb: Visualizes and quantifies the evolution of network communities over time. Supports the Network Analysis (Part 1 methodology of the [report]).
• results_analysis.ipynb:
• Evaluates the performance of the community-aware portfolio using cumulative returns, Sharpe ratios, volatility, and drawdowns. Supports the portfolio evaluation strategy and *evaluation (*Part 2 methodology section of the [report]).

Author

• Sonia Borsi (Dataism Laboratory of Quantitative Finance, Virginia Tech)