A Study Of Stock Selection Strategies Based On Random Forest Model With The Style Rotation Between Large And Small Caps

With the rapid development of computer technology and financial technology,quantitative investment has attracted attention in the stock market.The market share of quantitative products in domestic and foreign markets is also expanding,but compared with the development process of quantitative investment at home and abroad,there is still a large space for progress in China's quantitative investment market.As a representative technology in artificial intelligence,machine learning methods are also widely used in the field of quantitative investment.Random forest algorithm,as a classification algorithm,has the advantages of easy data processing,easy to understand parameters and results,and is suitable for quantitative research.Combining the advantages of machine learning algorithms,seeking the application of machine learning algorithms in the field of quantitative investment,it's of great significance for expanding market investment channels and promoting the development of the stock market.The paper uses the Shanghai Stock Exchange 50 constituent stocks and China Securities 500 constituent stocks as research samples from 2013 to 2018.After verifying the phenomenon of large and small disk style rotation in China's A-share market,the applicability of the quantitative investment method combining the idea of large and small disk style rotation and multi-factor stock selection in China's stock market was studied using the random forest method.The paper uses the Shanghai 50 and China 500 indexes to construct quantitative index of large and small cap styles,and selects financial factors,technical factors and macro factors as characteristic factors for multi-factor stock selection research.After comparing the fit of three machine learning models of logistic regression,support vector machine,and random forest,monthly frequency rolling strategies are constructed,combined with the actual investment performance of the strategy in China's Ashare market,the validity and robustness of the random forest stock selection strategy based on the style rotation between large and small caps were explored.Based on this,the contribution of the three types of characteristic factors in quantitative stock selection is further analyzed by excluding financial factors,technical factors and macro factors.The empirical results in this paper show that the random forest model in the field of quantified stock selection shows stable and excellent performance,can effectively identify the complex pattern between feature factors and excess returns,and has achieved comparative logistic regression and support in strategy Vector machine model for better investment performance.Secondly,combining the ideas of large and small cap styles of rotating stock selection and multi-factor machine learning stock selection can effectively improve investment performance and obtain 37% annualized returns.Finally,in the factor contribution analysis,it is found that among the three types of factors,the predicted contribution of financial factors is the highest,and the predicted contribution of macro factors is the lowest.But after excluding the macro factors,the annualized return of the strategy decreased by 7.5%,which shows that the macro factors also have predictive power in the quantitative stock selection strategy.