Multi-factor Quantitative Stock Option Planning Based On XGBoost Algorithm

In recent years,quantitative investment has been paid more and more attention by institutional investors and hedge funds with its characteristics of discipline,systematization,timeliness and decentralization.At the same time,the scale of China’s securities investment market and the number of securities accounts are increasing rapidly.From the experience of the securities market and the development of foreign securities market,the development prospect of quantitative investment is undoubtedly worth to be expected in the future.Nevertheless,the current domestic quantitative investment products still exist the overall size of small,quantitative strategy of a single,strategic performance differentiation and other shortcomings.At this point,the study of new ways to quantify investment and mining the importance of new modeling ideas naturally self-evident,for the rich quantitative investment products,enhance market size,to promote the development of quantitative investment is significant.In many quantitative strategies,multi-factor stock selection strategy by virtue of its stability and wide coverage and other advantages by many researchers attention.Multi-factor stock selection strategy is mainly focused on solving the multi-factor selection is comprehensive,followed by the classification model has a good generalization ability,based on these two directions,this article has carried out a certain optimization and improvement,one of the first A total of 307 factors were used in addition to the financial,bonus,momentum and other factors used by most researchers.I also added scale,valuation,macro,bond and property related factors.Using the more novel XGBoost lifting algorithm,the main advantage of this algorithm is that XGBoost supports linear classifiers and comes with logical regression or linear regression of L1 and L2 regularization terms.Second,XGBoost adds a regular term to the cost function to make the model model more simple and prevent overfitting.Finally,XGBoost draws on random forest practices,supports column sampling,not only reduces over-fitting,but also reduces computations,And XGBoost tools support parallel,faster.And compared the advantages and disadvantages of SVM,random forest and XGBoost.The results show that the XGBoost algorithm has the best effect and stability.Thirdly,this paper changes the previous factor filtering and modeling process.Training side screening method,the screening method is more scientific and reasonable.Based on the above planning ideas,the final successful design of the use of machine learning methods to quantify the stock picking,and achieved beyond the CSI 300 index excess rate of multi-factor quantitative stock selection program,after 23 holdings of the selected stock The combined yield of 287%,the annual compound yield of up to 127%,Sharp ratio of 0.91,the information ratio of 2.41,82% quarter outperformed the Shanghai and Shenzhen 300 Index,59% of the quarter to obtain positive returns,the final net Reaching 3.87,far exceeding the benchmark CSI 300 Index yield.