A Deep Learning-Based Model for Stock Price Forecasting Leveraging Investor Sentiment

Abstract

Abstract: The research presents the MS-SSA-LSTM model, which integrates multi-source data, sentiment analysis, swarm intelligence algorithms, and deep learning techniques to enhance stock price predictions. This model incorporates sentiment analysis from East Money forum posts, creating a unique sentiment dictionary and calculating a sentiment index. This offers valuable insights into market sentiment's influence on stock prices. The Sparrow Search Algorithm (SSA) is used to fine-tune LSTM hyperparameters, optimizing prediction accuracy. • Experimental results showcase the MS-SSA-LSTM model's superior performance. It's a valuable tool for accurate stock price predictions. Tailored for China's volatile financial market, the model excels in short-term stock price predictions, offering insights for dynamic decision-making by investors.And also, a hybrid LSTM+GRU model was introduced for stock sentiment classification. Additionally, a robust ensemble strategy was adopted, incorporating a Voting Classifier (AdaBoost + RandomForest) for sentiment analysis and a Voting Regressor (LinearRegression + RandomForestRegressor + KNeighborsRegressor) for stock price prediction. These ensembles seamlessly integrated with existing models (MLP, CNN, LSTM, MS-LSTM, MS-SSA-LSTM), collectively enhancing overall predictive performance. To facilitate user interaction and testing, a user-friendly Flask framework with SQLite support was developed, streamlining signup, signin, and model evaluation processes.