Multi-Factor Quantitative Trading Strategy Design Based On Random Forest Algorithm

In the case that domestic market investors increasingly favor quantitative trading investment with discipline and other advantages,this paper takes the theory of multi-factor model as the support,and builds multi-factor quantitative trading strategy on the basis of random forest algorithm.This paper divides the strategy into two parts:factor selection and quantitative stock selection.In this paper,it is believed that the traditional strategy can be improved by using the random forest algorithm in the factor selection and quantitative stock selection,so the multi-factor scoring strategy and multi-factor regression strategy based on the traditional method are made,and the double-random-forest strategy is made on the basis of the traditional multi-factor strategy.Among them,the multi-factor scoring strategy applies method of the information coefficient in the factor screening part and the rating method in the quantitative stock selection part.Multi-factor regression strategy applies the sorting method in the factor screening part and the regression method in the quantitative stock selection part.In the double-random-forest strategy,the random forest algorithm is used in both factor selection and stock selection.At last,we compare them in factor screening and result analysis of backtesting.The strategy of this paper is to take the constituent stocks of CSI 300 index as the stock pool.In the multi-factor model,factor analysis is especially important,so this paper will add two factors selected from the literature to the factor pool to enrich the factor pool.In this paper,we also verify the benefits of the two selected literature factors in the single-factor strategy.In the factor selection part,based on the annual data between April 2014 and April 2019,the strategies of this paper construct the factor library data based on the operating factors,valuation factors,growth factors and factors from the literature,and use the random forest algorithm,the method of information coefficient and the sorting method to screen the factors.In the quantitative stock selection part,based on the annual data between April 2015 and December 2019,the strategies ofthis paper apply the random forest algorithm,scoring method and regression method to select stocks.From the results of the backtesting,the double-random-forest strategy is suitable to be held in a long period of time.From April 2015 to December 2019,by comparing the three strategies in different periods,the overall annualized return of the dual-random forest strategy was 21.81%,surpassing the annualized return of the multi-factor scoring strategy,multi-factor regression strategy and CSI 300 index.From the perspective of risk control,the double-random-forest strategy is more suitable for holding in the periods of small rise,small fall and big rise.Therefore,by applying the random forest algorithm in the factor selection part of the strategy and the quantitative stock selection part,the double-random-forest strategy in this paper really improves the overall performance of the strategy to a certain extent.