Price Forecast And Comparative Study Of Stock Index Futures Based On Machine Learning Algorithms

With the opening of the capital market,the opening of financial derivatives,including stock index futures,is also highly anticipated.How to predict future price changes will be one of the important issues that investors will care about.Based on literature research and theoretical analysis,this article firstly selects the future price and volume trading information,technical indicators,and major market index price and volume information of all trading days in 2019 as input features.The research sample is the CSI300 stock index futures,and 1-minute high-frequency data is selected.On the basis of the decision tree,random forest,XGBoost,BP neural network and support vector machine algorithms,five price prediction models were built.Updating each model and optimizing the corresponding hyper-parameters of models at the end of each month,the paper predicts the future price and compares and analyzes the performances of all models.The trading strategy of "buying when price is going up and selling when down" according to the forecast result is designed and the back-test is carried out.Finally,as the robustness test,this paper also selects data from all trading days in 2018 and 2017 to repeat the above operation.All the results show that the support vector machine algorithm has the lowest error,high accuracy and stability in the prediction of stock index futures price.In addition,the trading strategy constructed based on this algorithm has a higher cumulative return and more controllable risk under certain risks,which is more suitable for stock index futures price prediction and investment application.