Wavelet transformation and predictability of Gold Price Index Series with ARMA model

Abstract

The U.S. gold futures market has recently attracted significant attention globally in the highly volatile equity and commodity futures markets. This study investigates an efficient algorithm based on ARMA denoising with wavelet transformation to measure the predictability of COMEX gold prices. The wavelet denoising decomposes and extracts the complex underlying structure and can reduce distortions occurring in the time series. The study has analyzed the COMEX gold time series for a period of the past five years, 2017-2022. The results show the outcome of alternative measures of predictability of the time series. The predictive measure with the traditional approaches assumes that the time series are linear and stationary over the long run and fails to explain the accuracy requirement in the short horizons. The results show a significant performance change compared to the conventional forecasting techniques.