Hybrid Deep Learning Approaches for Stock Price Prediction Using LSTM and GRU Models

Our study combines Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) recurrent neural networks to create an advanced stock price prediction model with superior time-series forecasting capabilities. Our research uses a complete stock price history coupled with SMA, EMA, RSI, and MACD technical indicators to strengthen the accuracy of our forecasting models. We prepare the data through MinMaxScaler normalization and generate time-delayed features to analyze stock price patterns. We set up two separate networks for LSTM and GRU before adjusting their parameters to help them perform better. We increase prediction accuracy by using a linear regression model to merge outputs from both LSTM and GRU systems. Our hybrid model performance is evaluated systematically through comparisons to individual LSTM and GRU results using both RMSE and R-squared metrics. In contrast to the standalone models, our hybrid design demonstrates superior performance by accurately forecasting stock price trends with an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value of 1.00 and an RMSE of 223.10. This research shows that combining the hybrid model provides stock market forecasters with a strong practical tool to help them develop better forecasts and investment strategies. Our results need future research to apply them to financial practice and develop new AI-based tools for market analysis.