Quantitative Investment Model Based On Improved GBDT

In recent years,quantitative investment has grown rapidly.It has become one of the main research directions of investment theories and strategies with its advantages of discipline and accuracy.With the development of quantitative investment in China,how to use the machine learning algorithm and stock data to establish a quantitative investment model to predict the future trend of the stock market is one of the main research directions of domestic quantitative investment.At present,domestic quantitative investment mainly involves two problems in predicting the future trend of stocks,choosing computer algorithms and establishing mathematical models.The research work of this paper includes two parts:five-factor model based on GBDT combination algorithm and trend-tracking model based on RF-GB algorithm.In this paper,a five-factor model with universal applicability in logic is established by constructing a GBDT combinatorial algorithm.In stock trend forecasting,the validity of the multifactor model has been confirmed.Combining the improved GBDT algorithm with the multi-factor model,the stock factor is firstly screened by the contribution degree and correlation analysis,and the multifactor model is constructed by obtaining five optimal factors,and then the GBDT algorithm is used to predict the stock trend.The empirical analysis shows that the five-factor model constructed based on the GBDT combination algorithm can effectively predict the future trend of stocks.This paper first constructs the RF-GB algorithm based on the random forest algorithm and the GBDT algorithm.The RF-GB algorithm improves the performance of the GBDT algorithm.The empirical results show that the RF-GB algorithm has better prediction accuracy than a single random forest algorithm and a single GBDT algorithm.Create a new combined trend tracking model,back testing by RF-GB algorithm shows that the new trend tracking model has a combination of advantages.