Research On Prediction Of Chinese Stock Index Futures Based On Machine Learning

Since its establishment in 2006,China Financial Futures Exchange has successively launched CSI 300 stock index futures,SSE 50 stock index futures,and CSI 500 stock index futures.While improving my country’s financial market,it has given full play to the function of promoting economic development.For the traditional financial market,the two-way trading mechanism of stock index futures can not only balance the market’s long and short forces and ensure the smooth operation of the market,but also help different market entities to hedge their value and establish a hedging mechanism to reduce price risks.A common concern of scholars and investors.When faced with complex financial data processing,traditional financial market forecasting methods are difficult to solve the problem of high noise in financial time series,and the nonlinear data characteristics also limit the performance of traditional methods in price forecasting.Machine learning represented by neural network has great advantages in data feature processing,data classification and regression.With the rapid development of computers,more concise,accurate and easy-to-use machine learning models can help investors tap more financial data features and effectively improve forecast accuracy.This paper takes the three major stock index futures of the China Gold Exchange as the research object,selects the 1-minute high-frequency data of all trading days from October 8,2021 to December 17,2021,and uses the futures volume and price trading information and technical indicators as input features.Firstly,EViews is used to conduct stationarity test,VAR regression analysis and Granger causality test to analyze the correlation of relevant input factors.Then select the frontier TabNet model in the field of deep learning,analyze the feature importance of each input feature in machine learning,and compare it with the correlation analysis test results.Secondly,the TabNet model and the GRU neural network model are used to predict,and the results are compared and analyzed.Finally,the model results are selected to construct a quantitative timing strategy with stock index futures as the trading object,and the strategy is optimized based on the backtest results.The research conclusion shows that the predicted value and the real value tend to be consistent with the rising and falling prices,and the time-selective trading strategy constructed from this has a positive rate of return within the backtesting range.From the research effect,the prediction model needs to be further optimized to improve the prediction accuracy;from the application effect,the application effect of machine learning on stock index futures prediction in quantitative investment is better.