Deep Learning Model for Stock Prediction Incorporating Sentiment Factors

Deep learning was a machine learning method based on artificial neural networks that enabled the learning and abstraction of features through multi-level non-linear models, thus facilitating the resolution of complex pattern recognition and prediction problems. Stock forecasting was the analysis of historical data and market trends with the objective of predicting future stock price trends and associated potential risks. Conventional stock prediction techniques based on deep learning typically relied on structured information, such as stock price time series and fundamental data, as a foundation for analysis and prediction. It had been observed that there was a close connection between stock market returns and investor sentiment, which demonstrated a positive correlation. However, prediction through sentiment analysis was not currently widely available. In light of this, this study introduced sentiment factors into the training of three models, namely LSTM, GRU, and Informer. The results demonstrate that the prediction accuracy of the model with additional sentiment factors is significantly enhanced compared to the model using only stock price data. Furthermore, the Informer-Sentiment model exhibits the most promising prediction effect among the three models incorporating sentiment factors.