gold price prediction using a hybrid convolutional-recurrent neural network (cnn-gru)

Gold, as a highly valuable asset, experiences frequent price fluctuations due to economic, political, and supply-demand factors, making accurate forecasting essential for investors and market analysts. a precise prediction model can help identify optimal buying and selling opportunities while minimizing financial risks. this study aims to develop a hybrid predictive model by integrating convolutional neural networks (cnn) and gated recurrent units (gru) to enhance the accuracy of gold price forecasting. in this framework, cnn is employed to extract spatial features from historical price data, while gru captures temporal dependencies, ensuring a more refined prediction. gold price data from 2004 to 2023 was collected, preprocessed, and normalized before being divided into training and testing datasets. the proposed model was trained using this dataset to identify patterns and trends in gold price movements. additionally, the implementation of multi-cycle models in the proposed methodology resulted in a 22–48% improvement in prediction accuracy compared to baseline hybrid recurrent models (cnn-lstm and cnn-bilstm) implemented in this study. the experimental results demonstrate that the cnn-gru model outperforms these alternatives in terms of forecasting precision. moreover, the proposed hybrid approach exhibits strong generalization capabilities, making it applicable to other financial time series forecasting problems. these findings highlight the effectiveness of combining cnn and gru in predictive modeling, providing a valuable tool for investors and analysts in making informed financial decisions. the novelty of this study lies in the introduction of a new hybrid cnn-gru model, applied for the first time specifically for gold price forecasting.