Comparison of Linear Regression and Holt-Winters Methods for Gold Price Prediction

Abstract

Gold is one of the most stable investment instruments and is widely favored by both individual investors and businesses. However, gold prices are influenced by various economic factors and often fluctuate significantly, making pricing decisions more challenging. Toko Mas Jasa Sejahtera, a gold trading business, faces difficulties in determining appropriate selling prices due to this uncertainty. Therefore, an accurate prediction method is required to minimize pricing errors and support more informed decision-making. This study aims to design and implement a gold price prediction system using the Linear Regression and Holt-Winters methods while comparing the predictive accuracy of both approaches. Model performance was evaluated using Mean Absolute Error (MAE) as the primary indicator of prediction accuracy. The results show that both methods are capable of forecasting gold prices, although with different levels of accuracy. The Linear Regression method achieved an MAE of IDR 44,097, whereas the Holt-Winters method produced an MAE of IDR 305,984. The substantially lower MAE obtained by Linear Regression indicates that it provides more accurate predictions than the Holt-Winters method. Therefore, Linear Regression is recommended as the preferred approach for the gold price prediction system at Toko Mas Jasa Sejahtera.