SCR-Financial-Networks Documentation

A Python framework for analyzing financial networks using Spectral Coarse-Graining (SCG), Agent-Based Modeling (ABM), and Graph Neural Networks (GNN).

Overview

This project implements a hybrid approach combining Spectral Coarse-Graining with Agent-Based Modeling and temporal Graph Neural Networks to analyze interbank contagion dynamics among European banks.

Key capabilities:

  • Spectral Coarse-Graining: reduce complex interbank networks while preserving diffusion dynamics

  • Agent-Based Simulation: model individual bank behaviors under stress scenarios

  • Temporal GNN: predict spectral properties from graph-structured market data

  • Risk Metrics: Delta-CoVaR, MES, and SCG-based systemic risk scores

  • Interactive Dashboard: Dash UI with live GNN training and SCG-vs-Basel comparison

  • Real Market Data: yfinance stock data and ECB macro indicators

Getting Started

Indices and tables