Paper: arXiv 2512.17185

Authors: Sandeep Neela

Abstract

Financial crises emerge when structural vulnerabilities accumulate across sectors, markets, and investor behavior. Predicting these systemic transitions is challenging because they arise from evolving interactions between market participants, not isolated price movements alone. We present Systemic Risk Radar (SRR), a framework that models financial markets as multi-layer graphs to detect early signs of systemic fragility and crash-regime transitions. We evaluate SRR across three major crises: the Dot-com crash, the Global Financial Crisis, and the COVID-19 shock. Our experiments compare snapshot GNNs, a simplified temporal GNN prototype, and standard baselines (logistic regression and Random Forest). Results show that structural network information provides useful early-warning signals compared to feature-based models alone. This correlation-based instantiation of SRR demonstrates that graph-derived features capture meaningful changes in market structure during stress events. The findings motivate extending SRR with additional graph layers (sector/factor exposure, sentiment) and more expressive temporal architectures (LSTM/GRU or Transformer encoders) to better handle diverse crisis types.

Complexity vs Empirical Score

  • Math Complexity: 4.5/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper presents an interesting application of multi-layer graph neural networks for systemic risk detection. While the conceptual framework is novel, the empirical implementation in this draft is stripped back to a single correlation layer with a small universe (44 stocks), making the mathematical and experimental validation somewhat preliminary.

Research Flowchart

  flowchart TD
    A[Goal: Early Warning for Systemic Market Crashes] --> B[Data Inputs: Cross-Asset Prices & Market Features]
    B --> C[Construct Multi-Layer Financial Graphs]
    C --> D[Compute Structural Graph Metrics & Train GNNs]
    D --> E[Benchmark Models: Snapshot GNNs, Temporal GNN, LR, Random Forest]
    E --> F[Validation: Tested on Dot-com, GFC, and COVID-19 Crises]
    F --> G[Outcome: Network Structure Outperforms Baselines in Crash Detection]
    G --> H[Future Work: Add Sentiment Layers & Transformer Encoders]

Cited References (19)

  1. Debtrank: Too central to fail? Financial networks, the fed and systemic risk (2012)
  2. Hierarchical structure in financial markets (1999)
  3. Financial risk prediction via graph neural networks (2020)
  4. Temporal graph attention for systemic risk prediction (2019)
  5. Random matrix theory and financial correlations (2000)
  6. Systemic risk and stability in financial networks (2015)
  7. Systemic risk in banking ecosystems (2011)
  8. Network theory: The emergence of the science of structure (2016)
and 11 more
  1. The liquidity and contagion channel of systemic risk (2009)
  2. Systemic risk in large financial systems (2013)
  3. Contagion in financial networks (2015)
  4. Financial networks and contagion (2014)
  5. Contagion in financial networks (2010)
  6. Monotone comparative statics for the fragility of financial networks (2016)
  7. Semi-supervised classification with graph convolutional networks (2017)
  8. Graph attention networks (2018)
  9. A comprehensive survey on graph neural networks (2020)
  10. Stock Pattern Assistant (SPA): A Deterministic and Explainable Framework for Structural Price Run Extraction and Event Correlation in Equity Markets (2025) — link
  11. AIMM: An AI-Driven Multimodal Framework for Detecting Social-Media-Influenced Stock Market Manipulation (2025) — link

0 of 19 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.