The Autoregressive Integrated Moving Average (ARIMA) is a flexible, good, and simple linear model for forecasting and time series analysis. Some time series forecasting researches also propose the Artificial Neural Network (ANN) model as a substitute nonlinear forecasting model. The ARIMA model is good at capturing linear patterns, but the ANN model is effective at capturing nonlinear patterns.
ANN and ARIMA models were significantly used in the prediction of the Egyptian stock exchange. Both ANN and ARIMA can also be merged as a hybrid model to capitalize on the capabilities of both models in linear and nonlinear modeling. We use the hybrid model in this research to merge ARIMA models and ANN model (the Deep Neural Network with numerous hidden layers) The Egyptian Stock Exchange is the actual dataset used.
The initial comparison made between the experimented prediction models for the time horizons of 10 days, 20 days, 30 days, 40 days, and 50 days in advance using the datasets in this work. To assess performance, statistical measurements for instance mean squared error (MSE), reveal that the DNN-ARIMA hybrid model outperforms non-hybrid models in predicting the Egyptian Stock Exchange and is particularly effective in improving prediction accuracy