Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Stock Price Prediction Using a Hybrid LSTM-GNN Model: Integrating Time-Series and Graph-Based Analysis

Meet Satishbhai Sonani, Atta Badii, Armin Moin

TL;DR

Stock price prediction remains challenging due to non-linear dynamics and inter-stock dependencies. The paper presents a hybrid LSTM-GNN model that fuses temporal sequence learning with graph-based inter-stock relationships, trained using expanding window validation. Empirical results show the hybrid achieves a $MSE$ of $0.00144$, a 10.6% reduction over a standalone LSTM ($0.00161$) and outperforms linear, CNN, and dense baselines. The work highlights the practical potential for real-time trading and financial analysis, while noting computational and validation considerations for deployment and future extensions to additional instruments and data sources.

Abstract

This paper presents a novel hybrid model that integrates long-short-term memory (LSTM) networks and Graph Neural Networks (GNNs) to significantly enhance the accuracy of stock market predictions. The LSTM component adeptly captures temporal patterns in stock price data, effectively modeling the time series dynamics of financial markets. Concurrently, the GNN component leverages Pearson correlation and association analysis to model inter-stock relational data, capturing complex nonlinear polyadic dependencies influencing stock prices. The model is trained and evaluated using an expanding window validation approach, enabling continuous learning from increasing amounts of data and adaptation to evolving market conditions. Extensive experiments conducted on historical stock data demonstrate that our hybrid LSTM-GNN model achieves a mean square error (MSE) of 0.00144, representing a substantial reduction of 10.6% compared to the MSE of the standalone LSTM model of 0.00161. Furthermore, the hybrid model outperforms traditional and advanced benchmarks, including linear regression, convolutional neural networks (CNN), and dense networks. These compelling results underscore the significant potential of combining temporal and relational data through a hybrid approach, offering a powerful tool for real-time trading and financial analysis.

Stock Price Prediction Using a Hybrid LSTM-GNN Model: Integrating Time-Series and Graph-Based Analysis

TL;DR

Stock price prediction remains challenging due to non-linear dynamics and inter-stock dependencies. The paper presents a hybrid LSTM-GNN model that fuses temporal sequence learning with graph-based inter-stock relationships, trained using expanding window validation. Empirical results show the hybrid achieves a of , a 10.6% reduction over a standalone LSTM () and outperforms linear, CNN, and dense baselines. The work highlights the practical potential for real-time trading and financial analysis, while noting computational and validation considerations for deployment and future extensions to additional instruments and data sources.

Abstract

This paper presents a novel hybrid model that integrates long-short-term memory (LSTM) networks and Graph Neural Networks (GNNs) to significantly enhance the accuracy of stock market predictions. The LSTM component adeptly captures temporal patterns in stock price data, effectively modeling the time series dynamics of financial markets. Concurrently, the GNN component leverages Pearson correlation and association analysis to model inter-stock relational data, capturing complex nonlinear polyadic dependencies influencing stock prices. The model is trained and evaluated using an expanding window validation approach, enabling continuous learning from increasing amounts of data and adaptation to evolving market conditions. Extensive experiments conducted on historical stock data demonstrate that our hybrid LSTM-GNN model achieves a mean square error (MSE) of 0.00144, representing a substantial reduction of 10.6% compared to the MSE of the standalone LSTM model of 0.00161. Furthermore, the hybrid model outperforms traditional and advanced benchmarks, including linear regression, convolutional neural networks (CNN), and dense networks. These compelling results underscore the significant potential of combining temporal and relational data through a hybrid approach, offering a powerful tool for real-time trading and financial analysis.

Paper Structure

This paper contains 28 sections, 4 equations, 6 figures.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Introduction
  4. Data Acquisition and Challenges
  5. Machine Learning Approaches in Financial Forecasting
  6. Research Challenges
  7. Methodology
  8. Data Collection and Preprocessing
  9. Feature Engineering
  10. Data Preprocessing
  11. Graphical Representation of the Stock Network
  12. LSTM Component
  13. GNN Component
  14. Hybrid Model Integration
  15. Training Strategy
  16. ...and 13 more sections

Figures (6)

  • Figure 1: Normalised closing prices of a sample stock with its 50-day and 200-day moving averages.
  • Figure 2: Graphical representation of the stock network.
  • Figure 3: Expanding window approach visualisation worsnup2022
  • Figure 4: MSE values across all test days using the best parameter configuration.
  • Figure 5: Comparison of MSE values across different models.
  • ...and 1 more figures