| As China’s economy is developing rapidly,the financial policy system has been gradually improved,and a variety of investment targets have emerged in the financial market.Financial derivatives such as stocks,funds,bonds,futures and options are investment products available for investors to choose from.An increasing number of experts and scholars hope to use the deep network model to help investors obtain high returns while reducing investment risks due to the rise of deep learning.Aiming at the stock price prediction,this paper builds a stock price prediction model based on deep learning theory and stock expertise.The main research work and achievements consist of the following two aspects.(1)Construction of a generative adversarial neural network model combining self-attention mechanism and residual networkAiming at the time series prediction problem of stock price prediction,this paper proposes a stock price prediction model in conjunction with the generative adversarial neural networks with self-attention and residual networks,and selects reasonable model parameters through experiments to improve the prediction performance of the model.(2)Construction of stock price forecasting model based on feature selection of feature engineeringIn deep learning research,features and data have an important impact on the prediction effect of the model.Features related to stock prices can improve the predictive performance of the model whereas irrelevant features not only waste computing resources,but also reduce the predictive performance of the model.On the basis of the SAR-GAN model,feature engineering is used to select and reduce dimensions of various features such as stock market indicators and technical indicators,and to preprocess the feature data according to the different dimensions of financial data,so that the prediction error of the model can be further reduced.Finally,the prediction results of the stock price prediction model are displayed via Django framework and Echarts components for easy reference by investors.This paper selects the Shanghai Stock Exchange Index and several hot industry stocks in different markets,such as Kweichow Moutai,Tencent,Apple,etc.,for comparative experiments.The experimental results show that the SAR-GAN model and the feature selection processing method proposed in this paper can produce prediction results with good performance on various evaluation indicators and can effectively reduce the prediction error when applied to stock price prediction.It can be seen that this study has important theoretical research significance and application value. |
