Challenges of Stock Prediction Based on LSTM Neural Network

For a long time, many scholars and researchers have tried stock forecasting. Stock forecasting has always been the most concerned and challenged in time series forecasting. People have different opinions on whether the stock market can be accurately predicted. Some scholars believe that stocks cannot be predicted, while others believe that using LSTM for stock prediction has high accuracy. In this work, the author experimented with whether LSTM could accurately predict stocks and found hysteresis in the prediction results. The author believes that although the prediction error of LSTM is small, it cannot provide support for actual transactions due to the hysteresis. In the last part of this paper, the author provides possible solutions to solve the hysteresis problem. These results explain how to improve the usability of stock market predictions and put forward suggestions and directions for future development.