GARCH-LSTM for Stock Price Prediction Using Sentiment Analysis

research introduces a new hybrid model called the SF-GARCH-LSTM model, designed for stock price forecasting. It combines sentiment analysis, long short-term memory networks (LSTM), and generalized autoregressive conditional heteroskedasticity (GARCH) models. The model first utilizes Bidirectional Encoder Representations from Transformers (BERT) to classify stock review titles into positive and negative sentiments, generating a sentiment factor (SF). Then, GARCH parameters are calculated using multiple GARCH models based on historical stock prices. Finally, the LSTM model combines stock price data, sentiment factors, and GARCH parameters to predict future stock prices. Experiment results demonstrate that the new proposed SF-GARCH-LSTM model significantly improves prediction accuracy compared to LSTM, SF-LSTM, and GARCH-LSTM models, highlighting the importance of incorporating sentiment information into financial forecasting.