A Novel Hybrid Temporal Fusion Transformer Graph Neural Network Model for Stock Market Prediction

Forecasting stock prices remains a central challenge in financial modelling, as markets are influenced by complex interactions between macroeconomic and microeconomic factors, firm-level fundamentals, and market sentiment. This study evaluates the predictive performance of classical statistical models and advanced attention-based deep learning architectures for daily stock price forecasting. Using a dataset of major U.S. equities and Exchange Traded Funds covering 2012–2024, we compare traditional statistical approaches, Seasonal Autoregressive Integrated Moving Average (SARIMA) and Exponential Smoothing, with deep learning architectures such as the Temporal Fusion Transformer (TFT), and the novel TFT-Graph Neural Network (TFT-GNN) hybrid that incorporates relational information between assets. All models are assessed under consistent experimental conditions in terms of forecast accuracy, computational efficiency, and interpretability. Results indicate that whilst statistical models offer strong baselines with high stability and low computational cost, the TFT outperforms them in capturing short-term nonlinear dependencies. The hybrid TFT-GNN achieves the highest overall predictive accuracy, demonstrating that relational signals derived from inter-asset connections provide meaningful enhancements beyond traditional temporal and technical indicators. These findings underscore the benefits of integrating relational learning into temporal forecasting frameworks and highlight the continued relevance of statistical models as interpretable, efficient benchmarks for evaluating deep learning approaches in high-frequency financial prediction.