A Novel Forecasting Model for Islamic Stock Prices: Hybrid ARIMA LSTM with Linear Regression OLS Techniques

This study focuses on improving the accuracy of Islamic stock price forecasting by developing a hybrid ARIMA-LSTM model integrated with OLS linear regression. Accurate predictions are critical for effective investment decisions, and while ARIMA and LSTM models are individually effective, their combination specifically for Islamic stocks has not been sufficiently explored. This study aims to bridge this gap by leveraging ARIMA to capture seasonal and trend components, LSTM for non-linear and long-term patterns, and OLS regression to integrate predictions from both models. This hybrid approach is designed to provide more accurate and stable stock price forecasting. Model performance will be evaluated using metrics such as MSE, RMSE, MAE, and MAPE, with the results expected to make a significant contribution to the field of stock price forecasting, particularly in the context of Islamic finance. The hybrid ARIMA-LSTM model shows higher accuracy compared to the ARIMA and LSTM models individually, both on training data ($0.52 \%$ higher accuracy than ARIMA, and $7.02 \%$ higher than LSTM) and testing data $\mathbf{(6 . 1 3 \%}$ higher accuracy than ARIMA, and $\mathbf{1 6 . 6 5 \%}$ higher than LSTM)