Stock Price Prediction Based On Lstm And Bert

Abstract

Price movements in the stock market affect all aspects of the social economy, and forecasting stock prices is of great importance. Traditional stock forecasting models are based on statistical regression models, which are difficult to characterize the influential relationships between multiple variables and predict stock price trends with large errors. In recent years, with the development of neural networks, neural networks have become a common method for stock forecasting, which include Back Propagation (BP) neural network, Convolutional Neural Networks (CNN), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) neural network. However, most of the previous stock price prediction models only use the basic stock market data, ignoring the influence of stock market investor sentiment on stock prices. A new stock price prediction model is proposed to address the above problems. First, the investor sentiment before the stock opening is calculated by fine-tuning the BERT model, then the calculated investor sentiment and the basic stock quotation data are aggregated, and finally the LSTM model is used to predict the closing price of the next stock trading day. We validate the effectiveness of the model on a real dataset of three Chinese listed companies.