Building A hybrid Time Series Model Using ARDL With LSTM and GRU Models

Purpose: The aim of the research is to utilize a hybrid model that combines the linear model represented by Autoregressive Distributed Lag (ARDL) and the nonlinear model represented by deep learning models, such as Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). Theoretical Framework: The theoretical framework integrates the linear component represented by the Autoregressive Distributed Lag (ARDL) model and the nonlinear component represented by deep learning models, namely Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) To create hybrid models. Design/Methodology/Approach: The research methodology involves the use of EVIEWS 12 for analyzing standard data and integrating the ARDL model, as well as Python programming for building the proposed forecasting models. Weekly data from the Iraqi stock market, specifically from the banking and communications sectors spanning from 2017 to 2021, is utilized. The study compares the performance of the hybrid models ARDL_LSTM and ARDL_GRU with individual models using evaluation metrics such as root mean square error (RMSE) and mean absolute percentage error (MAPE). Findings: The results indicate the superiority of the hybrid model ARDL_LSTM over other models due to its high accuracy and lower comparison measurement values. Originality/Value: The originality of the research lies in its hybrid approach, combining ARDL with deep learning models like LSTM and GRU for time series forecasting. This approach adds value by addressing the limitations of individual models and improving forecasting accuracy in the context of the Iraqi stock market.