Price forecasting in the precious metal market : A multivariate EMD denoising approach

Research output: Journal Publications and Reviews (RGC: 21, 22, 62)21_Publication in refereed journalNot applicablepeer-review

9 Scopus Citations
View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Original languageEnglish
Pages (from-to)9-24
Journal / PublicationResources Policy
Volume54
Early online date7 Sep 2017
Publication statusPublished - Dec 2017

Abstract

The precious metal markets are subject to the influence of complicated factor characterized by the interrelationship and nonlinearity with the short burst of noise data components. In this paper we propose a new Multivariate Empirical Mode Decomposition (MEMD) denoising model to identify the noise factors in the multiscale domain and forecast the precious metal price movement. Since the MEMD model is introduced to analyze and project the inter-relationship between different precious metal prices in the multiscale domain, the transient noise factor is identified, analyzed and suppressed. The movement of the reconstructed precious metal price is modeled using the ARMA model with higher accuracy. Empirical studies using the typical precious metal price data show that the proposed model achieves the statistically significant forecasting performance improvement, which provides the ex-post evidence on the noise factors identified. Further comparative studies of both MEMD and wavelet analysis based models show the complimentary relationship between these two popular multi scale models. We also found that Gold and Silver markets are subject to the similar influence of disruptive noises while Palladium and Platinum markets are subject to the influence of other influencing factors. The disruptive influencing factor is expected to be Euro/Dollar exchange rate.

Research Area(s)

  • ARMA model, Error entropy minimization, Multivariate Empirical Mode Decomposition (MEMD), Precious metal markets, Wavelet analysis