Research On GBDT-SVM Multi-level Stock Selection Model Based On A Large Number Of Factors

In the field of quantitative investment,the multi-factor model is widely accepted and used by investors in A-share market because of its high stability and high capital capacity.But recent years,model homogeneity has been getting worse and the style of A-share market has been changing frequently;as a result,investment based on multi-factor model can hardly obtain a considerable return.At present,in the context of the widespread application of artificial intelligence methods and the rapid development of computing capabilities,sorting out the problems in the application of multi-factor stock selection model and combining machine learning algorithms under big data samples with multi-factor model will bring a method for the optimization of multi-factor investment model.This work presents GBDT-SVM multi-level model based on big factor database,to promote the multi-factor model's ability to acquire excess return in stock investment by optimizing factor selection and dynamic adjustment for factor weight using machine learning techniques.Based on the classical multi-factor model,this model uses the gradient boosting decision tree algorithm to construct feature combination of the stock selection factors at the first step,and then combines the features as samples and uses the support vector machine algorithm to construct the stock selection model at the next step.Then,this work conducts empirical research using Chinese A-share market data and compares the model with the classical multi-factor model and its improved version.The research results show that the GBDT-SVM multi-level stock selection model has higher stock classification accuracy and shows good profitability in the historical backtesting.The results of this work mainly include two aspects.On the one hand,this work constructs a large number of special factors based on high-frequency trading data,and also validates the effectiveness of these factors,providing a rich library of factors for the future research of the multi-factor model.On the other hand,the GBDT-SVM multi-level stock selection model proposed in this work gives a new scheme for the optimization of the multi-factor model,and can give advice for Chinese investors.