The integration of multiple forecasting models has been widely recognized in both theatrical and empirical research as an effective approach for improving the predictive performance of individual models in time series analysis. This improvement becomes particularly significant when the combined models capture distinct characteristics of the underlying data. Hybrid forecasting models, which decompose a time series into its linear and nonlinear components, have emerged as a one of the most popular and effective approaches for handling complex and volatile datasets. In this study, the daily prices of gold and silver are forecast using a hybrid modeling framework, artificial neural networks (ANN), and conventional Box-Jenkins time series models. To rigorously evaluate and compare the forecasting performance of these models, Friedman’s test and the Morgan-Granger-Newbold (MGN) test are employed. The empirical results demonstrate that the hybrid model consistently outperforms both the traditional Box-Jenkins and feed-forward neural network (FFNN) models in terms of forecasting accuracy across different datasets.