Research On The Trend Forecast Of Shanghai And Shenzhen 300 Index Based On EMD-XGBoost

When traditional statistical model analysis methods are applied to non-linear financial data,problems such as low calculation efficiency and insufficient prediction accuracy arise.With the development of machine learning,more machine learning algorithms are applied to the financial market to solve the above problems.Among many machine learning algorithms,GBDT(Gradient Boosting Decision Tree)is used as an improved XGBoost(e Xtreme Gradient Boosting)algorithm,which has the characteristics of high accuracy,low computational complexity,and fast running speed.Applying the XGBoost model to the field of financial securities markets can improve the accuracy of predictions and improve the efficiency of predictions at the same time.In addition,applying the empirical mode decomposition algorithm(EMD)which is commonly used in signal processing to the financial field is also very effective.Based on the above background,in this article,an enhanced forecasting algorithm called EMD-XGBoost has been proposed.The algorithm is applied to the prediction of the trend of the Shanghai and Shenzhen 300 Index on the next trading day.Firstly,important technical indicators containing data features based on the historical market data of the Shanghai and Shenzhen 300 Index are extracted;secondly,the Empirical Modal Decomposition Algorithm(EMD)is used to decompose the index's daily time-sharing data into trend component and volatility component so that the index's daily data is divided into trend or volatility by the calculated volatility energy ratio;thirdly,XGBoost model is established as well as its parameters are optimized by using the cross-validation method and the grid search method for the purpose of obtaining a better fitting effect of the stock price trend prediction model;finally,the EMD-XGBoost enhanced forecasting model is built to forecast the trend of the next trading day.The experimental results show that the model is better than decision tree model(C5.0)and support vector machine model(SVM)as well as XGBoost model in the classification accuracy of price trend prediction of CSI 300 index.