The Shanghai Composite Index Prediction Scheme Based On XGBoost Algorithm Design Research

Data mining technology generated in the late 1980 s,and had the development which progresses by leaps and bounds in 90 s,as the technology matures,more and more scholars used it widely in different areas;Among them,with the combination of the financial sector can bring additional benefits to the vast number of investors;The stock market is a complex system affected by various information,the stock market fluctuations due to its high instability,is hard to predict.Investors in the face of a large number of stock market information,usually want to be able to use the known historical information in some way to predict future market movements,for use in investment,gain excess returns.Face huge amount of information,artificial processing is clearly unrealistic:the cost was too expensive;So many scholars use machine learning methods such as support vector machine,BP neural network,to forecast stock market rise and fall;This area has gradually become the hot spot problems of nearly two years to solve;But the support vector machine(SVM)method has certain limitations,in order to achieve the optimal classification effect,have to adopt high latitudes plane,this undoubtedly increases the complexity of the model;XGBoost algorithm has been known from 2015,with the advantage of high computation efficiency and accuracy,became the proposed algorithm.So the author using this new algorithm to forecast the stock market up or down,to provide a new effective solution for the investors;Based on the domestic stock market and the main stock index,using support vector machines,decision tree model and XGBoost algorithm to predict the rise and fall for the Shanghai composite index,Shanghai 50 index,standard&poor's index;At the same time,as much as possible in order to improve the algorithm of support vector machines,decision tree and XGBoost forecasting effect of the stock market up or down,the author also adjusted data related with the volume data,make its value are not too big compared with other indicators;We also tuning the parameters of the XGBoost algorithm.Selected the 28 technical index as input variables,then predict the second day of the stock market rise and fall as output variables classification;Using RStudio software to support vector machines,decision tree and XGBoost modeling,and relatively reasonable empirical results are obtained,the results show that the XGBoost forecast for the Shanghai composite index has a very ideal effect,prediction accuracy reached more than 70%,This is associated with the principle of XGBoost algorithm,it iterative error each time,minimized square loss function,so the accuracy is higher than common algorithms;Shanghai 50 and Standard&Poor's index prediction accuracy reached 60%to 65%,this may be related to these two indexes selection is only part of the shares as sample;When carried out in accordance with the trend of division,also can get a higher forecasting accuracy,Using the predicted results based on XGBoost algorithm to invest and the results also show that investors can acquire ideal excess return,support vector machine and decision tree are a little low,also reached more than 60%.You can see that the machine learning method for stock market forecast and investment has certain guiding significance.This can help investors decision-making and government regulation meanwhile provides a convenient,practical and feasible solution.