Stock quantitative prediction analysis method based on deep learning transformer self-attention mechanism

In recent years, China's economy has developed rapidly, and accordingly, financial market has also developed rapidly in China, attracting the attention of investors domestic and foreign. Therefore, it is of vital significance to study the stock price trend in China's financial market for scholars, investors and regulators. With the rise of quantitative trading and other ideas, more and more scholars apply deep neural network (DNN) to the financial field. Although DNN has achieved great success in image, voice and text in recent years, it has encountered many challenges in financial time series prediction due to the highly dynamic and noisy feature of dataset. As a typical representative of DNN in time series data processing, LSTM, because this method does not consider the importance of data at different time points and different sources, the effect is still not ideal. Different from the introduction of the Attention mechanism on the traditional LSTM model, by improving the Self-Attention model, the daily data and the time-sharing data are encoded and fused separately, and the impact of changes in capital flow on changes in stock trends can be learned. The experimental results show that the proposed method improves the accuracy of trend judgment to 63.04%, and obtains a return of 6.562% in the two-month back testing experiment, which proves that the model has a certain effectiveness and practicability in stock price trend prediction.