Hybrid Evolutionary-Optimized Deep Learning Model (Heodl) For Gold Price Prediction

Abstract

Gold serves as a hedge against inflation and economic uncertainty, making it a crucial duty in financial markets to predict gold prices.  Since gold price changes are complicated, nonlinear, and volatile, traditional forecasting models frequently have difficulty keeping up.  We present the Hybrid Evolutionary-Optimized Deep Learning (HEODL) model, which combines Random Forest Feature Selection (RF-FS) for dimensionality reduction, Differential Evolution (DE) for hyperparameter optimization, and Long Short-Term Memory (LSTM) networks for sequential learning in order to overcome these difficulties.  Techniques from Explainable AI (SHAP) are also used to improve the interpretability and transparency of the model.  The results show that the HEODL model achieves greater accuracy in gold price forecasting, outperforming baseline models like ARIMA, conventional LSTM, and Random Forest. The use of explainability approaches also enables a better understanding of the primary economic variables influencing variations in the price of gold, making HEODL a useful tool for financial analysts, investors, and policymakers.