Research On Multi-factor Stock Selection Model Based On Random Forest Algorithm

The purpose of this paper is how to combine machine learning algorithm with traditional multi-factor stock selection model,and constructs a multi-factor stock selection model based on random forest algorithm,sorts stocks by random forest algorithm to select valuable stocks,and constructs an effective portfolio.This paper uses all A shares as a stock pool,a large class of factors as a factor pool,selects separately 23 factors of value class,growth class,momentum class,financial quality class,technology class and analyst emotion class as candidate factor,and selects data for the last trading day of each month from January 2010 to December 2017 as factor data,builds a sample set with factor data and corresponding next stock monthly yield.The sample from January 2010 to December 2013 is used for model parameters optimization,to determine the hyper-parameters of the random forest algorithm and the optimal training window length.The sample from January 2014 to December 2017 is used for out-sample model back-test,to analysis the performance of stock selection model.The multi-factor stock selection model based on random forest algorithm is a dynamic stock selection model,which is trained using sample data from the past 6 months during each back-test period,uses current factor data to make predictions,selects the top 50 stocks with the predicted probability as the next stock position,and allocates them with equal weight.The whole construction process of this model can be roughly divided into three parts: data preprocessing and effective factor screening,model parameter optimization and result analysis,model improvement and optimization.In this paper,the multi-factor stock selection model based on the random forest algorithm has a total return of 160.05% and annualized income of 27.64% during the back-test period from January 2014 to December 2017,and it significantly leads the market benchmarks(CSI 300 and CSI 500),which can be proved that the stock selection model has better stock picking performance.Compared with the non-dynamic learning model,the dynamic stock selection model in this paper reflects its timeliness and the changes of the market to a certain extent.In addition,in the improvement and optimization of the model,the performance of original stock selection model can be improved by weighting prediction probabilities to weights,using the factor importance to re-filter the factors and considering the rotation effect of factors.