Paper: arXiv 2510.04357

Authors: Anoushka Harit, Zhongtian Sun, Jongmin Yu

Abstract

We propose the Causal Sphere Hypergraph Transformer (CSHT), a novel architecture for interpretable financial time-series forecasting that unifies \emph{“Granger-causal hypergraph structure”}, \emph{“Riemannian geometry”}, and \emph{“causally masked Transformer attention”}. CSHT models the directional influence of financial news and sentiment on asset returns by extracting multivariate Granger-causal dependencies, which are encoded as directional hyperedges on the surface of a hypersphere. Attention is constrained via angular masks that preserve both temporal directionality and geometric consistency. Evaluated on S&P 500 data from 2018 to 2023, including the 2020 COVID-19 shock, CSHT consistently outperforms baselines across return prediction, regime classification, and top-asset ranking tasks. By enforcing predictive causal structure and embedding variables in a Riemannian manifold, CSHT delivers both \emph{“robust generalisation across market regimes”} and \emph{“transparent attribution pathways”} from macroeconomic events to stock-level responses. These results suggest that CSHT is a principled and practical solution for trustworthy financial forecasting under uncertainty.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts including Riemannian geometry, hypergraph theory, and causal inference with rigorous formulation, while also presenting empirical backtests on S&P 500 data across multiple tasks and market regimes.

Research Flowchart

  flowchart TD
  A["Research Goal: Forecast Asset Returns using Financial News & Sentiment"] --> B["Input Data: S&P 500 Returns & News (2018-2023)"]
  B --> C["Methodology: Causal Sphere Hypergraph Transformer<br/>- Granger-Causal Hypergraph Extraction<br/>- Riemannian Geometry Embedding<br/>- Causally Masked Transformer"]
  C --> D["Computation: Training on Market Regimes incl. COVID-19 Shock"]
  D --> E["Outcomes:<br/>- Superior Return Prediction & Regime Classification<br/>- Robust Generalization<br/>- Interpretable Causal Attribution"]

Cited References (114)

  1. ModAugNet: A new forecasting framework for stock market index value with an overfitting prevention LSTM module and a prediction LSTM module (2018)
  2. Contrastive learning with heterogeneous graph attention networks on short text classification (2022)
  3. Time series analysis: forecasting and control (2015)
  4. A LSTM-based method for stock returns prediction: A case study of China stock market (2015)
  5. Incorporating corporation relationship via graph convolutional neural networks for stock price prediction (2018)
  6. Enhancing Stock Movement Prediction with Adversarial Training. (2019)
  7. Money: Ensemble learning for stock price movement prediction via a convolutional network with adversarial hypergraph model (2023)
  8. A Rewiring Contrastive Patch PerformerMixer Framework for Graph Representation Learning (2023)
and 106 more
  1. A generative bayesian graph attention network for semi-supervised classification on scarce data (2021)
  2. Deepmove: Predicting human mobility with attentional recurrent networks (2018)
  3. Support vector machine with adaptive parameters in financial time series forecasting (2003)
  4. Learning phrase representations using RNN encoder-decoder for statistical machine translation (2014) — link
  5. Gated-attention readers for text comprehension (2016) — link
  6. Modeling extreme events in time series prediction (2019)
  7. Temporal relational ranking for stock prediction (2019)
  8. Using structured events to predict stock price movement: An empirical investigation (2014)
  9. Ml-gat: A multilevel graph attention model for stock prediction (2022)
  10. Stock selection via spatiotemporal hypergraph attention network: A learning to rank approach (2021)
  11. Hypergraph neural networks (2019)
  12. Multiple stocks recommendation: a spatio-temporal hypergraph learning approach (2024)
  13. A unified approach to interpreting model predictions (2017)
  14. Explainable AI (XAI) models applied to planning in financial markets (2021)
  15. Interpretable dropout prediction: towards XAI-based personalized intervention (2024)
  16. Finbert: Financial sentiment analysis with pre-trained language models (2019) — link
  17. Is unimodal bias always bad for visual question answering? a medical domain study with dynamic attention (2022)
  18. Adaptive distance message passing from the multi-relational edge view (2023)
  19. Breaking Down Financial News Impact: A Novel AI Approach with Geometric Hypergraphs (2024) — link
  20. Convolutional Differentiable Logic Gate Networks (2024) — link
  21. Graph representation learning in bioinformatics: trends, methods and applications (2022)
  22. Revisiting the role of heterophily in graph representation learning: An edge classification perspective (2023)
  23. Graph neural networks for graphs with heterophily: A survey (2022) — link
  24. Robust Graph Structure Learning under Heterophily (2024) — link
  25. Heterophily and graph neural networks: Past, present and future (2023)
  26. Interpretable neural-symbolic concept reasoning (2023)
  27. The pushshift reddit dataset (2020)
  28. DBLP Computer Science Bibliography: Monthly Snapshot Release of November 2019 (2019)
  29. The movielens datasets: History and context (2015)
  30. The dynamics of viral marketing (2007)
  31. Semi-supervised classification with graph convolutional networks (2016) — link
  32. Graph attention networks (2017)
  33. Hypergcn: A new method for training graph convolutional networks on hypergraphs (2019)
  34. Neural similarity learning (2019)
  35. Graphs and Hypergraphs. Graphs and Hypergraphs (1973)
  36. Hypergraph learning: Methods and practices (2020)
  37. Geometric deep learning: going beyond euclidean data (2017)
  38. Sphere Neural-Networks for Rational Reasoning (2024) — link
  39. Hypergraph Mining for Social Networks
  40. Matrix factorization techniques for recommender systems (2009)
  41. Learning with hypergraphs: Clustering, classification, and embedding (2006)
  42. Hyperbolic recommender systems (2018) — link
  43. Deep learning based recommender system: A survey and new perspectives (2019)
  44. Spherical and hyperbolic embeddings of data (2014)
  45. Robustness, Heterogeneity and Structure Capturing for Graph Representation Learning and its Application (2024)
  46. Csrec: Rethinking sequential recommendation from a causal perspective (2024) — link
  47. Poincar 'e embeddings for learning hierarchical representations (2017)
  48. Neural matrix factorization++ based recommendation system (2021)
  49. Mind: A large-scale dataset for news recommendation (2020)
  50. Goodbooks-10k: a new dataset for book recommendations (2017)
  51. Improving recommendation lists through topic diversification (2005)
  52. Avicenna: a challenge dataset for natural language generation toward commonsense syllogistic reasoning (2022)
  53. Neural collaborative filtering (2017)
  54. Lightgcn: Simplifying and powering graph convolution network for recommendation (2020)
  55. Kgat: Knowledge graph attention network for recommendation (2019)
  56. Neural Reasoning for Sure Through Constructing Explainable Models (2025)
  57. Causal disentangled variational auto-encoder for preference understanding in recommendation (2023)
  58. HSR: Hyperbolic social recommender (2022)
  59. Monitoring Behavioral Changes Using Spatiotemporal Graphs: A Case Study on the StudentLife Dataset (2024)
  60. xdeepfm: Combining explicit and implicit feature interactions for recommender systems (2018)
  61. Explicit factor models for explainable recommendation based on phrase-level sentiment analysis (2014)
  62. " Why should i trust you?" Explaining the predictions of any classifier (2016)
  63. Trustsvd: Collaborative filtering with both the explicit and implicit influence of user trust and of item ratings (2015)
  64. Geoopt: Riemannian optimization in pytorch (2020) — link
  65. Sentence-bert: Sentence embeddings using siamese bert-networks (2019) — link
  66. Deep learning with long short-term memory networks for financial market predictions (2018)
  67. The graph-based behavior-aware recommendation for interactive news (2022)
  68. Explainable recommendation: A survey and new perspectives (2020)
  69. Causal embeddings for recommendation (2018)
  70. Stock Recommendations for Individual Investors: A Temporal Graph Network Approach with Mean-Variance Efficient Sampling (2024)
  71. Community detection on networks with Ricci flow (2019)
  72. Ricci curvature of the internet topology (2015)
  73. Returns to Buying Winners and Selling Losers: Implications for (1993)
  74. FinMME: Benchmark Dataset for Financial Multi-Modal Reasoning Evaluation (2025) — link
  75. Financial sentiment analysis using FinBERT with application in predicting stock movement — → 2306.02136
  76. Deep causal reasoning for recommendations (2024)
  77. Self-supervised learning for time series analysis: Taxonomy, progress, and prospects (2024)
  78. Dynamic graph neural networks for enhanced volatility prediction in financial markets — → 2410.16858
  79. Causal data science for financial stress testing (2018)
  80. Stress testing network reconstruction via graphical causal model (2021)
  81. Meta-transfer learning for low-resource abstractive summarization (2021)
  82. A model of investor sentiment (1998)
  83. Giving content to investor sentiment: The role of media in the stock market (2007)
  84. Attention is all you need (2017)
  85. Detecting and quantifying causal associations in large nonlinear time series datasets (2019)
  86. Investigating causal relations by econometric models and cross-spectral methods (1969)
  87. Twitter mood predicts the stock market (2011)
  88. Sentiment analysis on social media for stock movement prediction (2015)
  89. Stock movement prediction from tweets and historical prices (2018)
  90. Hypergraph convolution and hypergraph attention (2021)
  91. Low-dimensional hyperbolic knowledge graph embeddings (2020) — link
  92. Forecasting stock prices from the limit order book using convolutional neural networks (2017)
  93. FinGAT: Financial graph attention networks for recommending top-$ k $ k profitable stocks (2021)
  94. Higher Order Transformers: Enhancing Stock Movement Prediction On Multimodal Time-Series Data — → 2412.10540
  95. Deepar: Probabilistic forecasting with autoregressive recurrent networks (2017) — link
  96. Heterogeneous hypergraph representation learning for link prediction (2024)
  97. A novel multi-source information-fusion predictive framework based on deep neural networks for accuracy enhancement in stock market prediction (2021)
  98. Transformer-based attention network for stock movement prediction (2022)
  99. RicciFlowRec: A Geometric Root Cause Recommender Using Ricci Curvature on Financial Graphs (2025)
  100. GLANCE: Graph Logic Attention Network with Cluster Enhancement for Heterophilous Graph Representation Learning (2025) — link
  101. TextFold: A Geometric Hypergraph Framework for Protein Structure Prediction with Scientific Literature Integration (2025)
  102. Actionable Interpretability via Causal Hypergraphs: Unravelling Batch Size Effects in Deep Learning (2025) — link
  103. Causal Spherical Hypergraph Networks for Modelling Social Uncertainty (2025) — link
  104. ManifoldMind: Dynamic Hyperbolic Reasoning for Trustworthy Recommendations (2025) — link
  105. Spar-gnn: Knowledge tracing with behavioural patterns and selective llm feedback (2025)
  106. Advanced hypergraph mining for web applications using sphere neural networks (2025)

3 of 114 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.