Gold Prices Forecasting Using Recurrent Neural Network with Attention Tuned by Metaheuristics

A large number of statistical studies have proven that the value of gold is highly valued nowadays, despite the numerous crises that have hit the global market. Due to the great risks that daily affect sudden changes in gold prices on the global market, it is of great benefit to develop tools that will enable automated monitoring and prediction of gold price changes. This can be quite helpful for those countries and major corporations whose economy depends on this. In this paper recurrent neural network with attention layer was used to forecast the gold price, while the hyperparameters of the network were tuned by the novel variant of the particle swarm optimization algorithm. Obtained results in terms of standard regression metrics were compared to those generated by other cutting-edge methods and proposed method proved to be very promising in this area.