TIME SERIES FORECASTING FOR PREDICTING LONG-TERM GOLD PRICE

Abstract

The accurate prediction of gold price is a crucial task for investment decisions and has a considerable impact on investors, policymakers, financial Institutes and countries that depend on mineral resources, yet it remains a significant challenge due to its inherent volatility and complex, non-linear dynamics. Therefore, understanding the future price movement is important for revenue-based planning for investors, companies and also the country. This research focuses on applying Machine Learning techniques for predicting the Gold Price and providing a comparative analysis of the 4 ML algorithms for long-term gold price forecasting namely SARIMA, ARIMA, LINEAR REGRESSION, SVM (kernel used are RBF, Linear and Polynomial). Utilizing a comprehensive dataset of gold prices from August 2000 to August 2025 which was pre-processed by resampling into monthly averages to model long-team trends. All the models were trained on the historical monthly data and employed to generate a 60-month (5-year) forecast, from September 2025 to August 2030.