The use of ARIMA, LSTM and GRU models in time series hybridization with practical application

The importance of forecasting has emerged in the economic field in order to achieve economic growth, as forecasting is one of the important topics in the analysis of time series, and accurate forecasting of time series is one of the most important challenges in which we seek to make the best decision. The aim of the research is to suggest the use of hybrid models for forecasting the daily crude oil prices as the hybrid model consists of integrating the linear component, which represents Box Jenkins models and the non-linear component, which represents one of the methods of artificial intelligence, which is long short term memory (LSTM) and the gated recurrent unit (GRU) which represents deep learning models. It was found that the proposed hybrid models in the prediction process when conducting simulation for different sample sizes and when applied to the daily crude oil price time series data, were more efficient than the single models, and the comparison between the single models and the proposed hybrid models was made by comparison scale, mean square error (MSE), and the results showed that the proposed hybrid models have the ability to predict crude oil prices, as they gave more accurate and efficient results.

Keywords:

ARIMA , Long Short Term Memory , Time Series Forecasting , Gated recurrent unit , Hybrid Model