Research On Event-Driven Stock Index Futures Trading Strategy Based On LSTM Model

China's securities market has a short history of development.Compared with developed countries,the development of securities market is imperfect and immature.One prominent feature is that small and medium-sized investors account for the majority,resulting in a large number of irrational investments.But the emergence of quantitative investment provides a good tool to solve such problems.This paper proves that the change of stock price in the short term is predictable.As we know,due to the complexity of various influencing factors and the ambiguity of influencing mechanism,it is extremely difficult to truly realize the completely accurate prediction of stock price,but for the prediction of stock index price,the neural network of machine learning has its application.This supervised learning simulates the human brain's memory pattern and can abstract the description of the random-like time series process.LSTM neural network is the latest model of neural network.Its selective memory and internal influence of time series are very suitable for the time series of stock price,which is a kind of random non-stationary series.Therefore,the author believes that the expansion and improvement of LSTM neural network will emerge in an endless stream in the future.This thesis mainly studies the BP neural network,RNN neural network and LSTM neural network technology application in the short-term stock price forecast,and trying to build a multiple time scale based LSTM quantitative investment strategies,and tentatively for new type of neural network technology is applied to the prediction of the stock market analysis to provide certain theoretical and practical value.The results show that the model has a broad application prospect in the prediction of csi 300 index with macro market comments and other variables.The conclusions of this paper mainly include :(1)the neural network has certain advantages in analyzing and forecasting the stock data with complex and significant non-linearity.Compared with the immature weak effective market like the Chinese market,the neural network has a good prediction ability.(2)aiming at the problems of the traditional financial time series model,the LSTM financial time series prediction model is proposed,and compared with the traditional BP neural network method,this prediction model has obvious advantages.(3)the LSTM financial time series prediction model used in this paper,when forecasting the usage price data and the macro market comment index,has a better prediction effect than the usage price data only,indicating that the macro market comment index can promote the market forecast.(4)on the basis of introducing the prediction model of neural network,we creatively applied it to the pattern classification of the securities market to judge whether the index is in the "up" or "down" from small level to large level,so as to build a three-dimensional time-space coupling strategy(TTC neural network model)based on LSTM.The original 3d time and space coupling theory can automatically screen unilateral market and guarantee relatively large profit and loss ratio and win rate.The innovation points of this paper are mainly reflected in the following three aspects :(1)according to the characteristics of massive information data in the securities market,chaos and unclear occurrence time point,the macro market comment index is constructed to solve the problems of cumbersome manual data analysis,subjective analysis results,quantitative analysis and inaccurate analysis.(2)research and implement a stock prediction method based on LSTM neural network.Firstly,it can collect stock price trend data and macro market comments,and model the influencing factors of stocks to predict stock price trend.(3)combining market sentiment change data,securities market data and financial timing prediction model,this paper presents a quantitative stock selection model and strategy based on lstm neural network in multiple time scales,proposes quantitative trading means,and measures the feasibility of the strategy.Inevitably,due to the limited research level,this paper still has some deficiencies :(1)when constructing the macro market comment index,it adopts the method of crawling the key words of the editorial,compiling the dictionary manually,and assigning key words,which is subjective.(2)there is a huge amount of information in the securities market,which covers many aspects of stocks,such as the background,operation,financial status,profitability,debt paying ability and corporate governance structure of listed companies.However,LSTM financial time series model is adopted in this paper.In terms of data,only quantitative price data and macro market comment data are selected,failing to cover all aspects of information.