Prediction Of Stock Closing Price Based On PCA And Seq2Seq Learning Model

Stock is the most important part of the financial market.It is a complex non-linear dynamic system affected by many factors.It has become a subject that scholars have been studying for a long time.Accurate and stable stock forecasting model is very important for investors to avoid risks and make profitable investment strategies.Therefore,stock forecasting model has important research significance.With the rapid development of computer technology and the arrival of big data era,neural network machine learning and deep learning technology have been greatly developed.Because of its strong nonlinear approximation ability and self-learning ability,it is widely used in time series prediction,which makes the research of stock ticket enter a new stage.Among them,cyclic neural network has outstanding performance in processing serialized data,with memory function,can save historical information,and can update with the input of new data,and can effectively solve the long-term and short-term dependence of time series.In this paper,the Seq2 Seq learning model in the cyclic neural network is used to predict the short-term closing price of stocks,and compared with a variety of financial time series forecasting models to explore the feasibility and applicability of Seq2 Seq model in the short-term trend prediction of stocks.The research data selected in this paper are two representative stock indexes in China's stock market,i.e.Shanghai Composite Index and Shenzhen composite index,to verify the universality of Seq2 Seq model stock forecast,and then select six activestocks with large market value and top trading volume in 2018 to verify the forecast effect.The selected daily data variables include date,opening price,closing price,lowest price,highest price,trading volume,turnover,turnover rate,up and down,up and down range.Firstly,PCA is used to reduce the dimension of the input sample data,which can not only avoid the redundancy of data,but also reduce the dimension of data and improve the efficiency of calculation.The Seq2 Seq model is used to predict the short-term trend of China's stock price.Through theoretical research and empirical comparative analysis,it is shown that Seq2 Seq model is suitable for stock time series prediction direction,and in the effect prediction of individual stocks,compared with a variety of commonly used stock time series prediction models,it can be found that the measurement accuracy of Seq2 Seq model remains at a high level,with excellent short-term trend prediction ability of financial time series.The innovation of this paper lies in the application of deep learning technology in the traditional financial time series prediction,which provides a new idea for the stock research,and gets a better prediction effect in the empirical research,which proves the feasibility of deep learning in the financial time series prediction,and can better guide the investment behavior of institutions and individuals.