ATTENTION-DRIVEN GRU MODEL WITH FEATURE FUSION FOR ROBUST STOCK PRICE PREDICTION
Abstract
In the realm of stock price prediction, an integration of advanced Deep Learning architectures with comprehensive data sources has become imperative for achiev- ing heightened accuracy. This study proposes an innovative approach leveraging an attention-driven Gated Recurrent Unit (GRU) model, enhanced by feature fusion techniques, to predict stock prices. The model incorporates historical price data and technical indicators, capitalizing on their collective insights into market trends and patterns. The attention mechanism embedded within the GRU archi- tecture dynamically emphasizes salient features from the input sequence, which allows the GRU model to focus on the most relevant information for prediction. Furthermore, the integration of feature fusion enables the seamless combina- tion of diverse data streams, facilitating a more holistic understanding of market dynamics. Through extensive experimentation and validation on real-world finan- cial datasets, our proposed model demonstrates superior predictive performance compared to baseline models, showcasing its efficacy in capturing intricate mar- ket behaviors and enhancing robustness in stock price forecasting. This research study contributes to an enhancement of predictive analysis in financial markets, offering a promising avenue for informed decision-making and risk management strategies.