This study successfully ﬁtted an artiﬁcial neural network to a series on gold prices. The data used was monthly gold prices in US dollars and cents per troy ounce from October 2004 to February 2020. Of the 17 suggested Artiﬁcial Neural Network structures, the one with 2, 6 and 1 neurons in the input, hidden and output layers (ANN (2-6-1)) was adjudged the best because it had the least error, Mean Square Error (MSE) and Mean Absolute Error (MAE). The adequacy of the selected model was further conﬁrmed by graphical examination of the actual values of gold prices and the ones predicted by the model as well as graphical residual analysis. Consequently, forecasts were made using the chosen network. The forecasts suggest a decline in gold prices in the coming months.
Keywords: Gold, Forecasting, ANN.