Gold Price Forecasting Using LSTM, Bi-LSTM and GRU

Abstract

Due to the multifactorial and non-linear nature of the gold market, it is difficult to predict the gold price. The gold price is affected by many external factors, such as market environment, economic crises, oil price increases, tax advantages and interest rates. Therefore, multivariate models can better predict the gold price than univariate models. This study investigated the effects of gold price, crude oil price, consumer price index, exchange rate index, stock market index, and interest indicators between 2001 and 2021. Models created using LSTM, Bi-LSTM and GRU methods were evaluated using lowest Root Mean Square Error (RMSE), Mean Absolute Percent Error (MAPE) and Mean Absolute Error (MAE) metrics. The LSTM model performed best, with 3.48 MAPE, 61,728 RMSE and 48.85 MAE values.