Stock Price Prediction Using Wavelet Transform and Neural Networks

Abstract

The study investigates the use of hybrid CNN+LSTM models for stock price prediction. The article examines methods for stock price prediction using wavelet transform and hybrid neural networks, specifically CNN+LSTM models. The study demonstrates that wavelet transform, specifically Daubechies wavelets, effectively extracts key trends in financial time series by removing noise and short-term fluctuations. CNN is utilized to identify local patterns, while LSTM captures long-term dependencies, ensuring high prediction accuracy. The hybrid CNN+LSTM model integrates these approaches, enhancing the analysis of complex financial data. This improved the quality of the input data for the model. The findings indicate a significant improvement in prediction accuracy compared to traditional methods. The research results open perspectives for further studies in the application of hybrid neural network models in financial forecasting. In particular, exploring other combinations of neural networks and data preprocessing methods is possible. Implementing this approach can be beneficial for financial analysts and investors, aiding in more informed investment decisions and reducing risks in financial markets. The study confirmed that hybrid models combining different types of neural networks are effective for complex tasks, such as financial time series forecasting.