Comparison of the two hybrid models, Wavelet-ARIMA and Wavelet-ES, to predict the prices of the US dollar index

Abstract

The US dollar index is one of the most important measures to compare the value of the US dollar against a basket of foreign currencies. The strategic importance of this index lies in avoiding risks and fluctuations in the basket of major global currencies. It is known that the process of accurate prediction must take place after understanding the nature of the data of the phenomenon under study, and accordingly we can employ the most appropriate models to obtain the best predictive values. In this paper, we made a comparison between two models from the hybrid wavelet transform models, namely Wavelet-ARIMA and Wavelet-ES, by applying to data representing the weekly rates of the last price of the US dollar index from 2011 to 2022, in order to get the best predictive values for this indicator. The results of the comparison criteria AIC, RMSE and MAPE indicated the preference of the hybrid Wavelet-ARIMA model, which was used to predict the weekly rates of the index (USDX). These results indicated that there would be no significant changes or fluctuations during the next sixteen weeks, the weekly average of the index price will be ($96), the lowest predictive value of the index will be ($95.24), which will be recorded in the fourteenth week, and that the fifteenth week will record the highest predictive value of the index, as it will amount to ($96.31).