Time Series Forecasting Enhanced by IntegratingGRU and N-BEATS
Milind Kolambe, Sandhya Arora · International Journal of Information Engineering and Electronic Business · 2025
Accurate stock price prediction is crucial for financial markets, where investors and analysts forecast future prices to support informed decision-making.In this study, various methods for integrating two advanced time series prediction models, Gated Recurrent Unit (GRU) and Neural Basis Expansion Analysis Time Series Forecasting (N-BEATS), are explored to enhance stock price prediction accuracy.GRU is recognized for its ability to capture temporal dependencies in sequential data, while N-BEATS is known for handling complex trends and seasonality components.Several integration techniques, including feature fusion, residual learning, Ensemble learning and hybrid modeling, are proposed to leverage the strengths of both models and improve forecasting performance.These methods are evaluated on datasets of ten stocks from the S&P 500, with some exhibiting strong seasonal or cyclic patterns and others lacking such characteristics.Results demonstrate that the integrated models consistently outperform individual models.Feature selection, including the integration of technical indicators, is employed during data processing to further improve prediction accuracy.