Gold Price Forecasting Comparison: Study of the Box-Jenkins Method and the Holt’s Method in United States of America

Abstract

Gold production in USA is much greater than other regions of the world. It has been estimated that gold production of USA is second largest in the history. The price of gold in USA might be affected by the macroeconomic factors and thus the gold prices must be forecasted in advance so that the accurate and relatable policies can be designed by the policy makers of USA. In this context, this study is conducted to estimate the accuracy of three models used for gold price forecasting i.e. Hot Winters, Box Jenkins and ANN for USA. The accuracy of these methods has been estimated through the specific criteria such as MSE, MAE and MAPE. Time series data for 20 years has been collected regarding gold prices in USA and has been taken in two subsections i.e. monthly and yearly. It was found out that Holt Winters is the most accurate method for the forecasting purposes and especially in the forecast of gold prices and can be used to predict the future values for several years. After Holt Winters, the second accurate method was found to be ANN or artificial neural network that presented lower MAPE values as compared to Box Jenkins method.