This study aims to enhance stock price prediction accuracy by integrating a gated recurrent unit (GRU) with an artificial rabbit optimization (ARO) algorithm. The objective is to address the issues in hyperparameter optimization and deliver a high-performance predictive model for stock market trends tested on the Dow Jones Industrial Average (DJIA) dataset.
The proposed ARO-GRU hybrid model uses a GRU for time-series stock price prediction and an ARO to dynamically optimize the model’s parameters. ARO-GRU was benchmarked against various models, including single-layer and multi-layer GRU, BiLSTM and long short-term memory (LSTM) models optimized by genetic algorithms (GA) or ARO. Performance was assessed using metrics such as the mean square error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and R-squared (R2).
The experimental results showed that the ARO-GRU model significantly outperformed its counterparts. Compared to the best alternative model (LSTM-ARO), ARO-GRU reduced the MSE by 81.8% (from 22.731 to 1.864 for the AAPL stock) and the MAPE by 64% (from 0.025 to 0.009). It achieved an average R2 score improvement of 5.3% across all tested stocks, demonstrating a better model fit. In addition, the ARO-GRU model required 83% less computational time than the LSTM-ARO model, further validating its efficiency.
This study introduces the integration of the ARO algorithm with the GRU for stock market prediction, marking a novel combination of efficiency and optimization. By demonstrating significant improvements in prediction accuracy and computation time, this study provides a robust and scalable solution for dynamic stock-trading systems.
