Research On Stock Quantitative Multi-factor Selection Based On Boosting Algorithm

Quantitative investment is a modern trading method that uses mathematical models and computer technology to practice investment concepts and investment methods to obtain stable excess returns.It is an active investment strategy whose development source can be traced back to 40 years ago,and reached its peak in Western countries in the 1990 s.Its investment performance is stable,its market size and share continue to expand,and it has been recognized by more and more investors.In the field of quantitative investment,multi-factor models are deeply loved by academia and the industry due to their simplicity,efficiency,and flexibility.Its core idea lies in the various factors that affect stock returns.The historical data of these factors is used to dig out effective factors.Construct an investment strategy,and determine the investment portfolio based on the quantitative model constructed by the strategy,and use the active investment portfolio method to obtain stable excess returns.With the development of machine learning,various machine learning algorithms have also been applied to multi-factor stock selection,and multi-factor stock selection strategies have been continuously updated and developed accordingly.This article takes the constituent stocks of the Shanghai and Shenzhen 300 index as the research object of the stock pool,selects 244 quantitative factors in the quantitative factor library of the Uqer platform as the initial candidate factors,and uses the regression method and the ranking method to screen the candidate factors respectively.There are 19 effective factors.For these 19 effective factors,the traditional scoring method,Ada Boost,GBDT and XGBoost classification algorithms and regression algorithms are used to construct a multi-factor stock selection model.The results show that both the Boosting algorithm and the traditional multi-factor scoring model can obtain excess returns that exceed the Shanghai and Shenzhen 300 Index,and the overall performance of the Boosting algorithm is better than the traditional multi-factor stock selection model,especially the GBDT regression algorithm and The XGBoost regression algorithm has a better performance,indicating that it is feasible to use the Boosting algorithm to improve the traditional multi-factor stock selection model.