An improved wavelet–ARIMA approach for forecasting metal prices

Abstract

Metal price forecasts support estimates of future profits from metal exploration and mining and inform purchasing, selling and other day-to-day activities in the metals industry. Past research has shown that cyclical behaviour is a dominant characteristic of metal prices. Wavelet analysis enables to capture this cyclicality by decomposing a time series into its frequency and time domain. This study assesses the usefulness of an improved combined wavelet-autoregressive integrated moving average (ARIMA) approach for forecasting monthly prices of aluminium, copper, lead and zinc. The performance of ARIMA models in forecasting metal prices is demonstrated to be increased substantially through a wavelet-based multiresolution analysis (MRA) prior to ARIMA model fitting. The approach demonstrated in this paper is novel because it identifies the optimal combination of the wavelet transform type, wavelet function and the number of decomposition levels used in the MRA and thereby increases the forecast accuracy significantly. The results showed that, on average, the proposed framework has the potential to increase the accuracy of one month ahead forecasts by $53/t for aluminium, $126/t for copper, $50/t for lead and $51/t for zinc, relative to classic ARIMA models. This highlights the importance of taking into account cyclicality when forecasting metal prices.