Employment the black box models to forecast the central bank’s foreign currency sales

Abstract

The research is aims to forecasting multi-variable time series using black box models that link the input series with the output series with a mathematical model as it includes two types of models, which are equation error models and output error models, where the model rank was determined using a number of statistical and engineering criteria, namely (AIC, AICC, BIC, LOSS, FPE, FIT) and choosing the model corresponding to the lowest values ​​of the criteria as the best model for forecasting the future values ​​of the Central Bank's sales of foreign currency as a series of outputs and the demand gap as an input series, The results of the analysis showed that the appropriate model for the data is the model ARMAX(1,2,3,1) By relying on the ARMAX model, the central bank’s sales were predicted for the next four months, and the forecast results were consistent with the original time series values, indicating the ARMAX model’s efficiency.

Keywords

• time series
• black box
• prediction error
• delay time