Predictive Research Of Stock Market Based On Financial Time Series And Deep Learning

Currently,with economic globalization and financial integration,the stock market is becoming increasingly complicated and presents many deviations which cannot be explained by classical financial analysis.However,at the same time,some classical financial statistical characteristics have striking similarities.This shows that although the stock market is complex,there are universal laws,and the operating laws behind it can be found through data mining.This paper,through sorting out the existing literature and current situation,aims to clarify the development tendency of China's A-share market,find out its potential contributing factors,and construct investor expectation index and measure investor sentiment on the basis of traditional financial analysis.Next,we select comprehensive index,classification index,sample index to conduct empirical analysis,and conduct predictive research on the volatility of China's stock market,the trend of bulls and bears,and the closing price of predictive research.First,the forecast of stock market volatility.Using ARIMA,ARCH,GARCH and other financial time series models,this paper makes a predictive research on the stock price fluctuation and trend of China's A share market.The ARMA model is used to fit the linear financial time series,and the GARCH model is used to fit the nonlinear time series of residuals.The results show that the ARIMA model performs well in the timeseries data with obvious trend items,but it performs poorly in the nonlinear financial time series with strong randomness and many disturbance items.The GARCH model has a significant fitting effect on the nonlinear sequence and has a clear description of the stock market volatility agglomeration phenomenon.Then stock market bull-bear trend forecast is conducted.Machine learning algorithm is used to make predictive analysis on the trend movement of stock market bulls and bears.The research shows that the ensemble learning based on weight voting idea has a high accuracy for the prediction of stock market bulls and bears,among which the prediction accuracy of XGBoost is up to 96%.In addition,the effect of neural network model is also very good,the accuracy of which is over 90%.Finally,the stock market price prediction is carried out in this paper based on the deep neural network LSTM model to predict the stock closing price.The results show that the prediction of the LSTM model for the stock closing price is similar to the actual trend,and the loss function value of the model decreases and gradually converges with the increase of the number of iterations.The model fitting effect is good.Based on the empirical analysis,this paper puts forward the prospect with a view to the trade war and COVID-19,and puts forward some suggestions from the perspectives of market regulation,information symmetry,investment structure,financial innovation.