A hybrid regularized GRU-LSTM model for enhanced stock price prediction: integrating memory efficiency and predictive accuracy

Stock price forecasting remains an open problem since the market dynamics are not linear, especially in today’s world. This paper has presented the Hybrid Regularized GRU-LSTM model to combine the short term predictability of GRU and the memory keeping capability of LSTM for better predictive power. The primary goal of the model is to fix flaws that the current stock prediction model with the use of the regularization techniques to enhance the performance by providing agility and flexibility in the volatile stock market. Suggested method covers the data preprocessing, selection of model architecture, model training process, and evaluation of the desired biomarker model. When comparing the results with ARIMA, LSTM, and GRU models, the proposed model reaches the highest level of accuracy and equals to 91.3%. Therefore, the two metrics of RMSE and MAE, the two types of errors support better handling showing the method is more lucrative to apply in short-term stock price prediction. The findings further affirm that the method utilized in amalgamating the unbounded memory of LSTM with the efficient computation ability of GRU results to an improved performance of the developed Hybrid GRU-LSTM model perfect for financial time-series forecasting.