Stock Prediction System Based on Bi-directional LSTM Model

Abstract

In the rapid development of China's socialist market economy, stock investment has become a focal point of public attention, and the accuracy of stock market predictions is crucial for investors. This study employs a Bidirectional Long Short-Term Memory (Bi-LSTM) network model to conduct an in-depth predictive analysis of a representative Chinese stock. The research initially selected stock data from 2010 to 2020, which underwent meticulous cleaning and normalization processes, and was prepared for model training through feature engineering and data segmentation. Subsequently, a Bi-LSTM model was constructed and trained, and its performance was evaluated using a validation set. Ultimately, the model's predictive capability was comprehensively assessed through key indicators such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R2). The study demonstrates the high application potential of the Bi-LSTM model in stock market prediction.