An Unified Model for Forecasting of Stock Market Price Using Machine Learning Algorithms

This study presents a comparative analysis of deep learning models for stock price forecasting, including Backpropagation Neural Network (BPNN), Bidirectional Long Short-Term Memory (BiLSTM), and Bidirectional Gated Recurrent Unit (BiGRU). We investigate two distinct hybrid approaches to enhance predictive accuracy: a stacking ensemble model, which employs a meta-learner to optimally combine the predictions of the base models, and a novel Hybrid with Attention model. The latter integrates the three architectures into a single, end-to-end network, utilizing an attention mechanism to dynamically weight the importance of temporal features from the BiLSTM and BiGRU branches. The performance of all models is evaluated on historical stock data using key metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R-squared (R2). The results demonstrate that the hybrid models, particularly the attention-based architecture, significantly outperform the individual base models by leveraging their combined strengths and intelligently focusing on the most relevant sequential information. The findings highlight the efficacy of advanced hybrid models in improving the accuracy and robustness of financial time-series predictions.