RNN-Based Financial Time Series Prediction: Performance and Prospects

Abstract

In this study, we proposed a Recurrent Neural Network (RNN)-based method for financial time series prediction. The performance of the proposed RNN model was compared with traditional machine learning models, including Support Vector Machine (SVM) and Random Forest (RF), as well as deep learning models like Long Short-Term Memory (LSTM). Experimental results demonstrated that the RNN model outperformed the other models in all three evaluation metrics: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). This suggests that the RNN model can effectively capture the trends and fluctuations in financial time series, providing accurate predictions for future stock price movements. While LSTM is a more sophisticated deep learning model, RNN showed superior performance in capturing short-term dependencies in the data, while maintaining training efficiency. Additionally, the results revealed that RNN’s simpler structure still achieves competitive accuracy in the prediction of stock price movements, making it an ideal model for real-time market forecasting. The findings highlight the potential of RNN in financial applications and open avenues for further exploration with more complex architectures like LSTM or Transformer models to capture long-term dependencies in financial data.