Research On Short-Term Stock Market Price Trend Prediction Based On Deep Learning

As a key component of the financial industry,the stock market is a matter of national importance.Stock price trend forecasting is especially important to investors and traders.Accurate forecasting of stock price trends provides important information about the future movement of stock prices.This helps investors to develop better investment strategies,reduce risk and achieve higher returns.Traditional stock price trend forecasting methods rely on technical analysis and fundamental analysis,which are limited by market noise,non-linear relationships and other factors and perform generally in the stock market.Therefore,more and more researchers are choosing deep learning techniques to predict stock price trends.With adaptive and nonlinear modeling capabilities,deep learning is able to extract effective features from a large amount of historical data and predict future stock price trends.Changes in stock prices are influenced not only by historical prices,but also by factors such as company fundamentals and market macroeconomic factors.Therefore,this paper constructs a multi-source feature fusion system MFF based on a priori knowledge of finance.The system divides the features required for stock price trend prediction into four parts: basic features,technical features,news features and correlation features,and uses principal component analysis to fuse the features of these four features.After constructing the multi-source feature fusion system,this paper proposes the Stock-MFFWGAN short-term stock price trend forecasting model based on Wasserstein generative adversarial network.The model can make full use of the multi-source features to explore the potential connections between different data and achieve accurate short-term stock price trend prediction.Through ablation experiments,it is demonstrated that the multi-source feature fusion system proposed in this paper can significantly improve the accuracy of stock price trend prediction.Comparison experiments are conducted between traditional deep learning methods and the Stock-MFF-WGAN model,and the experimental results show that the Stock-MFF-WGAN model achieves optimal results on leading stocks in six different industries.Finally,the model is back-tested in this paper,and the experiments demonstrate that the Stock-MFF-WGAN model also outperforms other control models in a real market environment.Based on the constructed multi-source feature fusion system and the proposed Stock-MFF-WGAN model,this paper uses Vue.js and Spring Boot framework to develop a short-term stock market price trend forecasting system.The system consists of five main modules: data collection module,forecasting model module,model deployment module,user interface module and security module.The design and implementation of the entire system follows the software development process,starting with requirements analysis,then implementing the system design,and finally conducting software testing.The system is based on Web services and can easily provide investors with assistance in stock market analysis and decision making.