Research On Price Forecast Of Shanghai Composite Index Based On LSTM Model

With the continuous development of China's reform and opening up,the level of national economy is constantly improving,the important position of the stock market in economic operations has become increasingly prominent.As of June 2018,the total market value of China's stock market has reached 53.89 trillion yuan,about 65%of China's full-year GDP in 2017.The report of the 19th National Congress proposed"To deepen the reform of the financial system,enhance the financial services of the financial services,increase the proportion of direct financing,and promote the healthy development of multi-level capital markets.Improve the dual-pillar regulation framework of monetary policy and macro-prudential policies,and deepen market-oriented reform of interest rates and exchange rates.Improve the financial supervision system and hold the bottom line of systemic financial risks."As the most important direct financing channel,the stock market directly links investors and the real economy.It is the main part and typical form of China's capital market.Trying to establish a more accurate stock price forecasting model can not only guide investors'stock investment behavior,but also effectively improve the stock market's service entity economic ability,and can investigate and prevent financial risk fluctuations,maintain financial stability for the country,and strictly control systemic financial risks.It has important guiding and practical significance for the macro-control of the stock market.The article selects the Shanghai Securities Composite Index?Shanghai Composite Index?to characterize the overall trend of the Chinese stock market.The sample period is from January 2,2014 to May 31,2018,for a total of 1,076 trading days.Among them,using the data from January 2,2014 to September 6,2017 as a model training set,the daily closing index of the Shanghai Composite Index for the period from September 7,2017 to May 31,2018 is predicted.Based on a detailed review of the literature on stock price forecasting research,the article improves the existing research from two directions:First,based on the modern financial theory of stock price fluctuations,combined with the actual situation of the Chinese stock market,theoretically establishes three Factor prediction model:P_t+1=f?transaction information_t,investor concern and investor sentiment_t,monetary policy_t?.On this basis,considering the authenticity,reliability and comprehensiveness of the data,the opening index,the lowest index,the highest index,the closing index,the trading volume,the transaction amount,etc.are selected as the proxy indicators of historical trading information,and the turnover rate and financial and economics are selected.News emotions as a proxy indicator of investor sentiment,select stocks and users of the forums and snowball stocks as the proxy indicators of investor concern,select one-year deposit interest rate,Shanghai Interbank Interbank 7-day interest rate,USD against RMB As the proxy indicator of monetary policy,the exchange rate constructs the historical transaction information comprehensive index?SH index?,investor concern and investor sentiment comprehensive index?IS index?and monetary policy comprehensive index?MP index?through principal component analysis;Long-term short-term memory?LSTM?neural network technology is used to construct the LSTM-0model based on the closing index sequence.The LSTM-1 model is constructed based on three types of comprehensive indices.The LSTM-2 model is built based on all the original indicators.Super-parameter tuning,improve the predictive ability of each model,compare the three models on the test set,the average error rate of the LSTM-0,LSTM-1,and LSTM-2 models are 3.19%,1.65%,and 0.40%.The LSTM-2 model based on all the original indicators has a relatively better prediction effect.The empirical results show that the LSTM model has a good application prospect in the nonlinear and multivariate stock price forecasting problem,and investors'concerns and investor sentiment,monetary policy and so on have a certain degree of influence on the Chinese stock market.The establishment of the LSTM model after the impact of reasonable quantification can effectively improve the accuracy of stock price forecasting.