EnVoR: Multi-Granular Gold Price Prediction Using Ensemble Voting Regression

Price prediction and forecasting have been important topics in the financial world. Among all commodities, such as gold, silver, platinum and others, gold has been the prime and forefront commodity which decides the price movements of other commodities. This paper proposes a unique multi-granular gold price prediction methodology, called EnVoR, by leveraging voting based regression tree models and utilizing simple, cumulative and exponential moving average values as features. Our proposed EnVoR predicts gold prices of next day, next week, next month and next quarter seasons. Extensive experiments are conducted on the gold price daily data from December 2011 to June 2024 to predict the future prices at various granularity levels. Experimental results on the gold price data have confirmed that the proposed EnVoR methodology has given superior performance in terms of highest regression accuracy and the lowest error values for MSE, RMSE, MAE and MAPE for predicting next day, next week, next month and next quarter price values. The proposed EnVoR model can be used by financial institutions and stock market analysts for providing buy, sell or hold recommendations to their clients.