This paper proposes a hybrid approach to dollar exchange rate forecasting, wherein both types of forecasting models-linear and non-linear models-have been incorporated to improve the efficiency of predictions. The work essentially combined MISO ARX model with GARCH-X models within MISO ARX framework to enhance the data. These three estimation techniques, namely IV4, RELS-HF, and RELS-SE, could optimize the modeling performance of the MISO ARX model, whereas Quasi Maximum Likelihood Estimation (QMLE) was applied for GARCH-X models.

An evaluation metric such as the Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) was used for the comparison between model performances. The findings show that RELS-SE method is superior to other estimation methods for MISO ARX models, and the results also suggest that the best forecasting accuracy was obtained for the MISO ARX (1,5,3,5,3) - GARCH (1,2)-X model. It has been effective in hybridizing the exchange rate changes model for cash volatility predictions as the hybrid model excellent catch in capturing volatility. This grows importance for a hybrid model in financial forecasting, especially in turbulent markets.

It is valuable for policy makers, financial analysts, and economic researchers. It suggests that hybrid time series models could be used to improve the accuracy of exchange rate forecasting. Future research should include investigation into the combination of machine learning techniques with hybrid econometric models to further enhance prediction improve performance.