Forecasting the COMEX copper spot price by means of neural networks and ARIMA models

This paper examines the forecasting performance of ARIMA and two different kinds of artificial neural networks models (multilayer perceptron and Elman) using published data of copper spot prices from the New York Commodity Exchange, (COMEX). The empirical results obtained showed a better performance of both neural networks models over the ARIMA. The findings of this research are in line with some previous studies, which confirmed the superiority of neural networks over ARIMA models in relative research areas. •The performance of the ARIMA model is good enough in terms of forecasting.•The mean forecast error of back-propagation neural network is lower than ARIMA׳s.•The mean forecast error of the Elman recurrent neural network is the lowest.•The variance of the back-propagation neural network is lower than ARIMA.•The variance of the forecast error of the Elman neural network is the lowest.