Hybrid Machine Learning and Data Analytics Framework to Improve Prediction of the Stock Market

Abstract

This research is to develop a hybrid model that combines statistical models such as SARIMAX and deep neural networks such as RNN/LSTM to improve the accuracy of financial market forecasting. The research relies on analyzing temporal patterns in financial data and using machine learning techniques to improve forecasting accuracy. Previous studies that used statistical models and deep neural networks were reviewed and demonstrated the superiority of deep models, but they focused on only one model. Therefore, this research explores the combination of statistical models with neural networks to achieve more accurate forecasting. SARIMAX and RNN/LSTM models were applied to Egyptian stock market data, with models trained over ten periods. Performance was measured using AIC, BIC, and Log Likelihood criteria, and residual errors were analyzed using tests such as Ljung-Box and Jarque-Bera to ensure the accuracy of the results. The results showed that combining the SARIMAX model with RNN/LSTM significantly improves forecasting accuracy compared to traditional models such as ARIMA and SARIMA, while reducing standard errors. Data dissemination was also improved using the ADF test, expanding the use of data from multiple markets to achieve more accurate results. The research emphasizes the importance of combining statistical models with deep neural networks to improve forecasting in financial markets. The results demonstrate that this combination significantly improves forecast accuracy compared to using a single model alone.