Stock Price Prediction Using Lstm and Gru Models Based On Historical Data Analysis

Authors: Padala Sri Roshni, Dr. Goldi Soni, Dr. Poonam Mishra

Abstract: The stock market is inherently volatile and influenced by complex patterns that traditional statistical models often fail to capture effectively. This research presents a deep learning-based approach for predicting stock prices using Long Short-Term Memory (LSTM) neural networks. The model is trained on historical stock price data obtained from Yahoo Finance, with a specific focus on Google (GOOG). In addition to the LSTM model, this research also integrates a Gated Recurrent Unit (GRU) model to compare performance and evaluate the efficiency of different recurrent neural architectures in stock price prediction. Data preprocessing techniques such as normalization, moving averages, and sequence generation were applied to enhance model learning. The LSTM architecture was designed to handle temporal dependencies within financial time series data, using multiple layers and dropout regularization to prevent overfitting. The model's performance was evaluated using metrics like Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE), demonstrating reliable predictive capability. Visual comparisons of actual versus predicted stock prices further affirm the effectiveness of the model. This study highlights the potential of LSTM networks in stock price forecasting and contributes to the advancement of intelligent financial decision-making tools.

DOI: http://doi.org/10.5281/zenodo.17731292