Gold Price Forecasting Using Disaggregation of Time Series Data

Abstract

The price of gold plays a crucial role in shaping investment strategies and influencing financial markets. Financial institutions, policymakers, and investors rely heavily on accurate gold price forecasts, given the unique position of gold as a safe-haven asset for hedging and diversification. Recognizing the increasing importance of gold in the eyes of investors, it becomes imperative to employ the most suitable forecasting technique. This study employs the Naïve method, ARIMA, Double Exponential Smoothing and K-Nearest Neighbours algorithm to predict future gold prices. The primary objectives include constructing forecast models for gold prices using these methods, determining the best-performing model among them, and comparing their forecasting performances using metrics such as mean absolute percentage error (MAPE), mean absolute error (MAE), root mean squared error (RMSE), and mean forecast error (MFE). The findings reveal that the ARIMA (1,1,1) model outperforms the Naïve method, Double Exponential Smoothing and KNN boasting the lowest values for MAPE, MAE, RMSE, and MFE—122.242, 7.90%, 154.136, and 133.350, respectively. In summary, the ARIMA (1,1,1) model demonstrates superior performance, making it the most accurate model for forecasting future gold prices