The Prediction Of Shanghai And Shenzhen 300 Index Based On H-LSTM Model

The price trend of stocks directly affects the economic interests of investors,and also affects and reflects the country's macroeconomic policies.The stock index is a financial indicator for stock market changes.Investors rely on the rise and fall of the index to determine the trend of stock price movements.Therefore,the price of stock index is a hot spot and focus of scholars.and inverstor.The classical econometric model has great limitations on the linear paradigm.it is difficult to reflect the nonlinear characteristics of financial time series.Due to the Markov assumption of stochastic differential model,it generally ignores a lot of inportant information.Predicting stock price movements is a very difficult thing,It is of great practical and theoretical significance to find a suitable model method for predicting the trend of China's stock index.We adopt the combination of theoretical analysis and empirical test to predict the stock price trend.Taking the CSI300 as the research object and the data from January 2006 to March 2019 is selected.First,we proposed a hybrid information extraction network based on supervised learning.The hybrid information extractor transforms the original input variables into new variables by learning the output characteristics of Heston and ARIMA.Establishing a bridge between machine learning methods and Heston and ARIMA methods.Then,based on the CSI300 price index,we compare the results of H-LSTM,econometrics,and stochastic differential methods to index price forecasts to explore the time series model that best matches the movement of the Chinese stock index.Finally,We further reserch the short-term effects of macroeconomic factors,the fundamentals of the weight indicator and the technical indicators on the price of the CSI300.testing the robustness of long-short term memory neural networks on the prediction of index prices.Through empirical research,we found that:(1)The prediction effect of stock price by neural network method is better than that of econometric model and stochastic differential model.Therefore,the CSI300 has a variation law that is difficult to be explained by the linear model.(2)Adding mixed information extractor,using batch prediction,point-by-point pushback,and controlling the rise and fall can significantly improve the accuracy price forecasting of LSTM.(3)In the case of predicting the price of the CSI300,increasing the number of network input variables and reducing the time width of the forecast can increase the fitting degree of the LSTM model to the index price.(4)The hybrid information extractor can effectively extract the characteristics of the CSI300.When the input variables are trading indicators,technical indicators,constituent stocks operating indicators and macroeconomic indicators,the forecasting effect based on the hybrid information extractor is better than the principal component analysis,the sparse Autoencoder and the t-SNE.