Paper: arXiv 2408.05659

Authors: Nikolas Michael, Mihai Cucuringu, Sam Howison

Abstract

We propose a novel data-driven network framework for forecasting problems related to E-mini S&P 500 and CBOE Volatility Index futures, in which products with different expirations act as distinct nodes. We provide visual demonstrations of the correlation structures of these products in terms of their returns, realized volatility, and trading volume. The resulting networks offer insights into the contemporaneous movements across the different products, illustrating how inherently connected the movements of the future products belonging to these two classes are. These networks are further utilized by a multi-channel Graph Convolutional Network to enhance the predictive power of a Long Short-Term Memory network, allowing for the propagation of forecasts of highly correlated quantities, combining the temporal with the spatial aspect of the term structure.

Complexity vs Empirical Score

  • Math Complexity: 7.0/10
  • Empirical Rigor: 6.0/10 (capped from 8.5: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.5)

Why this score: The paper employs advanced machine learning architectures (GCN-LSTM) and graph theory to model complex, high-dimensional temporal-spatial dependencies, which involves non-trivial mathematical formulations. It is also heavily data-driven, with explicit details on data preprocessing, graph construction, hyperparameter tuning, multiple evaluation metrics, and backtesting on real financial data (ES-mini and VX futures).

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Forecast ES & VX Futures"] --> B["Data Sources<br>ES-mini & VX Futures"]
  B --> C["Network Construction<br>Correlation-based Graph"]
  C --> D["Spatial-Temporal Modeling<br>GCN-LSTM Hybrid"]
  D --> E{"Training & Validation"}
  E --> F["Key Findings"]
  
  subgraph F ["Outcomes"]
      F1["Enhanced Forecast Accuracy"]
      F2["Interpretable Term Structure"]
      F3["Capturing Cross-Market Dynamics"]
  end

  style A fill:#e1f5fe,stroke:#01579b,stroke-width:2px
  style F fill:#f1f8e9,stroke:#33691e,stroke-width:2px

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a novel data‑driven network framework that combines a multi‑channel Graph Convolutional Network with a Long Short‑Term Memory network for forecasting returns, volatility and volume of E‑mini S&P 500 and VIX futures.

Source sentence
We propose a novel data-driven network framework for forecasting problems related to E-mini S&P 500 and CBOE Volatility Index futures, in which products with different expirations act as distinct nodes.

Assumes

No explicit assumptions found in the text.

Data

Backtest sample: E-mini S&P 500 and CBOE Volatility Index futures · futures · intraday

Costs

Not modeled A trading or execution result is reported without costs.

Finds 4

Reported performance (GCN-LSTM): Sharpe 10.94

  • GCN‑LSTM achieves the best Mean Absolute Error for trading‑volume forecasting on 3 out of 4 SPX products and is best or second‑best for 5 out of 8 VIX products, outperforming LSTM on the most liquid contracts. performance
    Source sentence
    MAE, GCN‑LSTM is the best‑performing method for 3 out of 4 SPX products and is either the best or second‑best method for 5 out of 8 VIX products.
  • GCN‑LSTM outperforms the base LSTM model on 9 out of 11 products for QLIKE and on 8 out of 11 products for HMSE in volatility forecasting, indicating the propagation step improves performance. predictive-accuracy
    Source sentence
    Thus, as GCN‑LSTM seems to outperform LSTM for 9 out of 11 products in terms of QLIKE and for 8 out of 11 products in terms of HMSE, the propagation step seems to result in an improvement.
  • GCN‑LSTM is the highest‑performing method among all methods for return forecasting on ES 1 and ES 2, delivering higher Sharpe Ratios and PPD than baselines. performance
    Source sentence
    GCN‑LSTM is the highest‑performing method among all methods for ES 1 and ES 2.
  • The proposed architecture provides significant improvements in forecasting accuracy over the base LSTM model across returns, volatility and volume tasks. predictive-accuracy
    Source sentence
    We have demonstrated that the proposed GCN‑LSTM architecture provides significant improvements in forecasting accuracy over the base LSTM model.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 15 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (45)

  1. Exploring Graph Neural Networks for Stock Market Predictions with Rolling Window Analysis (2019) — link
  2. Determinants of price discovery in the VIX futures market (2017)
  3. Modelling and trading the U.S. implied volatility indices. Evidence from the VIX, VXN and VXD indices (2016) — link
  4. Long short-term memory (1997)
  5. Understanding the VIX ([]) — link
  6. Stock returns and volatility: An empirical investigation of the German and French equity markets (1996) — link
  7. Understanding the VIX (2009) — link
  8. A Model of Asset Trading Under the Assumption of Sequential Information Arrival (1976)
and 37 more
  1. Price dynamics in the regular and E-mini futures markets (2004)
  2. A Term Structure Model for VIX Futures (2013) — link
  3. Link Prediction Based on Graph Neural Networks (2018) — link
  4. Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting (2021) — link
  5. DropEdge: Towards Deep Graph Convolutional Networks on Node Classification (2020) — link
  6. Deep Learning for Portfolio Optimization (2020)
  7. Integrated GCN-LSTM stock prices movement prediction based on knowledge-incorporated graphs construction (2024) — link
  8. Price discovery in the S&P 500 index derivatives markets (2016) — link
  9. Graph Neural Networks for Anomaly Detection in Industrial Internet of Things (2021) — link
  10. Hybrid graph convolution neural network and branch-and-bound optimization for traffic flow forecasting (2023) — link
  11. Semi-Supervised Classification with Graph Convolutional Networks (2017) — link
  12. Fin-GAN: forecasting and classifying financial time series via generative adversarial networks (2024) — link
  13. The Graph Neural Network Model (2009) — link
  14. Research on Graph Neural Network in Stock Market (2022) — link
  15. Volatility components: The term structure dynamics of VIX futures (2010) — link
  16. A Subordinated Stochastic Process Model with Finite Variance for Speculative Prices (1973)
  17. VIX Index: CBOE Volatility Index Methodology (2024)
  18. Volatility forecast comparison using imperfect volatility proxies (2011) — link
  19. The Relationship between Trading Volume, Volatility and Stock Market Returns: A test of Mixed Distribution Hypothesis for A Pre- and Post Crisis on Kuala Lumpur Stock Exchange (2005) — link
  20. Option Volume Imbalance as a predictor for equity market returns (2022) — link
  21. Causality of price movements in VIX exchange-traded products and VIX futures contracts (2023) — link
  22. The lead–lag relation between VIX futures and SPX futures (2024) — link
  23. Neural networks and arbitrage in the VIX (2020) — link
  24. Forecasting returns in the VIX futures market (2019) — link
  25. The VIX futures basis: Evidence and trading strategies (2014)
  26. VIX futures (2006) — link
  27. Dynamic Volume-Return Relation of Individual Stocks (2002)
  28. Modeling the Stock Relation with Graph Network for Overnight Stock Movement Prediction (2020)
  29. Graph Neural Networks for Forecasting Multivariate Realized Volatility with Spillover Effects — → 2308.01419
  30. A Review on Graph Neural Network Methods in Financial Applications (2021) — link
  31. mGNN: Generalizing the Graph Neural Networks to the Multilayer Case (2021) — link
  32. A Survey on Graph Neural Networks for Time Series: Forecasting, Classification, Imputation, and Anomaly Detection (2023) — link
  33. Graph Representation Learning (2020)
  34. The Price Impact of Order Book Events (2013) — link
  35. A novel graph convolutional feature based convolutional neural network for stock trend prediction (2021) — link
  36. Forecasting Stock Prices Using Stock Correlation Graph: A Graph Convolutional Network Approach (2021) — link
  37. Chart GCN: Learning chart information with a graph convolutional network for stock movement prediction (2022) — link

1 of 45 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.