Hybrid Method of Forecasting Central Bank Sales of Dollars

By fusing the most effective aspects of machine learning and statistics, hybrid approaches hold the potential to enhance time series forecasting.The basic notion is that by merging Artificial Neural Networks (ANN) and the Autoregressive Integrated Moving Average (ARIMA) model, this combination makes up for the shortcomings of one strategy with the advantages of the other.The linear model, which can only model linear relationships, and the nonlinear model, which can only model nonlinear relationships, are the two types of time series that the hybrid model can handle.Time series data of the Central Bank of Iraq's sales of US dollars during the period from 2003 to 2024 were used.The accuracy measures, Mean Square Error (MSE) and Mean Absolute Percentage Error (MAPE) of the hybrid combination were also compared with ARIMA and ANN methods.The results showed a significant improvement in the MSE and MAPE values of the hybrid model.