GRU-PFG: Extract Inter-Stock Correlation from Stock Factors with Graph Neural Network

ArXiv ID: 2411.18997 “View on arXiv”

Authors: Unknown

Abstract

The complexity of stocks and industries presents challenges for stock prediction. Currently, stock prediction models can be divided into two categories. One category, represented by GRU and ALSTM, relies solely on stock factors for prediction, with limited effectiveness. The other category, represented by HIST and TRA, incorporates not only stock factors but also industry information, industry financial reports, public sentiment, and other inputs for prediction. The second category of models can capture correlations between stocks by introducing additional information, but the extra data is difficult to standardize and generalize. Considering the current state and limitations of these two types of models, this paper proposes the GRU-PFG (Project Factors into Graph) model. This model only takes stock factors as input and extracts inter-stock correlations using graph neural networks. It achieves prediction results that not only outperform the others models relies solely on stock factors, but also achieve comparable performance to the second category models. The experimental results show that on the CSI300 dataset, the IC of GRU-PFG is 0.134, outperforming HIST’s 0.131 and significantly surpassing GRU and Transformer, achieving results better than the second category models. Moreover as a model that relies solely on stock factors, it has greater potential for generalization.

Keywords: Graph Neural Networks (GNN), GRU, Stock Prediction, Factor Investing, Time Series Forecasting, Equities

Complexity vs Empirical Score

• Math Complexity: 6.5/10
• Empirical Rigor: 8.0/10
• Quadrant: Holy Grail
• Why: The paper employs Graph Neural Networks (GNNs) combined with GRUs, representing advanced machine learning architectures with non-trivial mathematical foundations. It also presents strong empirical results on a standard financial dataset (CSI300) with specific performance metrics (IC 0.134) and comparisons against established benchmarks, indicating a high degree of implementation and backtesting readiness.

  flowchart TD
    Start["Research Goal: Predict stock returns using only factors, capturing inter-stock correlations without extra data"] --> Inputs["Data Inputs: Stock Factors Historical Data"]
    Inputs --> Method["Core Methodology: GRU-PFG Model"]
    Method --> Step1["1. Feature Extraction: GRU processes temporal factor sequences"]
    Step1 --> Step2["2. Graph Construction: Project factors into correlation graph"]
    Step2 --> Step3["3. Graph Learning: GNN extracts inter-stock correlations"]
    Step3 --> Step4["4. Prediction: Combine temporal & correlation features for return forecast"]
    Step4 --> Outcomes["Key Findings"]
    Outcomes --> F1["IC Score: 0.134 on CSI300 dataset"]
    Outcomes --> F2["Outperforms GRU & Transformer, surpasses HIST (IC 0.131)"]
    Outcomes --> F3["Matches 2nd-category models while using only standard factors"]
    Outcomes --> F4["Superior generalization potential"]