Research On Trading Strategy Of Shanghai And Shenzhen 300 Stock Index Futures Based On XGBoost

Stock index is a price index which is designed to measure the trendency of the stock market's overall price level.It can reflect the economic changes sensitively,so it has the title of "barometer" of modern economy.As a result,stock index futures with stock index as trading object have huge investment value.Many domestic and foreign researchers have been trying to use various methods to predict the price change of stock index futures.XGBoost(eXtreme Gradient Boosting)is a relatively new machine learning algorithm with high operational efficiency and accuracy.Using this algorithm to predict the price of stock index futures can help investors to make investment decisions,and has strong theoretical and practical significance.On the basis of previous research on security price prediction model and literature review,this paper improves the prediction effect of XGBoost algorithm on stock index futures price by optimizing input vectors and adopting more scientific anti-over-fitting methods.Firstly,this paper makes an in-depth analysis of the trend of the Shanghai and Shenzhen 300 index in the sample interval,and establishes a factor pool of 46 technical indicators including four categories:trend type,oversold type,volume type and stop-loss type,so as to ensure that the information contained in the price of stock index futures can be fully excavated.Secondly,according to the importance score of each feature calculated by XGBoost model after training,the features with the higher importance ranking are selected as the input vector of the prediction model,which reduces the complexity of the model and improves the prediction effect of the model.In this paper,the traditional cross-validation method and the time series cross-validation method are distinguished from each other theoretically,and the reasons why the traditional cross validation method is liable to cause the model over-fitting are analyzed in depth.The empirical results show that compared with the traditional cross-validation method,the time series cross-validation method has more advantages in both the prediction effect and the operation speed of the model.Accordingly,this paper constructs the trading strategy of Shanghai and Shenzhen 300 stock index futures based on XGBoost,and compares the performance of the strategy outside the sample with the trading strategy based on Random forest model.The results show that the former achieves higher returns.Finally,this paper summarizes the shortcomings in building prediction model and counstructing investment strategy,and points out the direction of future research.