Stock market prediction remains one of the most challenging and consequential problems in computational finance. This paper presents a comprehensive deep learning framework leveraging Long Short-Term Memory (LSTM) recurrent neural networks for time-series forecasting of equity closing prices. Using five years of historical trading data (2013–2018) from NSE-listed Tata Global Beverages Limited, comprising 1,235 trading observations, we construct a stacked dual-layer LSTM architecture trained on a 60-day lookback window with MinMax normalization to prevent data leakage. Our model—trained over 5 epochs with a batch size of 2 and the Adam optimizer—achieves convergence with a Mean Squared Error (MSE) loss of approximately 9.06 × 10⁻⁴. On an 80/20 train-validation split (987/248 observations), the model demonstrates strong temporal alignment between predicted and actual closing prices. The experimental results validate LSTM's effectiveness in capturing long-range sequential dependencies in financial time-series data, outperforming traditional statistical models in non-stationary market environments. This work contributes a reproducible, modular pipeline for equity price forecasting with practical implications for algorithmic trading, portfolio management, and financial risk modeling.
Tyagi et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: