A hybrid CNN-LSTM-GRU system tuned by whale optimization algorithm for enhanced gold and brent price prediction

• We propose an integrative system based on CNN, LSTM, and GRU to forecast gold and Brent prices. • CNN and GRU are employed in parallel to respectively learn local features and trends and generate the initial forecasts. • GRU learns and aggregates forecasts from CNN and LSTM to produce the final price prediction. • WOA is adopted to tune the parameters of CNN, LSTM, and GRU. • The proposed integrative forecasting systems outperforms baseline prediction systems. Forecasting the price of gold and crude oil is essential for government, economic agents and investors. In this regard, an integrative system based on convolution neural network (CNN), long short-term memory (LSTM) neural network, and gate recurrent unit (GRU). In the first stage of the integrative system, CNN and LSTM are employed in parallel to respectively learn local features and trends in data to generate the initial forecasts. In the second stage, GRU learns and aggregates forecasts from CNN and LSTM to produce the final price prediction. In addition, whale optimization algorithm (WOA) is adopted to tune the parameters of CNN, LSTM, and GRU. When applied and tested on gold and Brent markets, the proposed CNN-LSTM-GRU-WOA outperformed competing predictive systems including CNN-WOA, LSTM-WOA, pruned regression trees (RT), k -nearest neighbors algorithm ( k NN), and nonlinear support vector regression (SVR). The simulation results provide valuable recommendations to policy makers, producers, and investors with respect to forecasting gold and Brent prices.