Stocks Selection Based On Multi-factor Model And Random Forest Algorithms

Quantitative investment relies on mathematical models to quantify investment strategies and investment logic,and it uses modern technology like computer technology as the means of realization.As a new investment method that has emerged globally in the past two decades,its development situation is unshakable.In 2004,China's first quantitative investment fund,China Everbright Prudential Quantitative Core Securities Investment Fund was established.In 2010,stock index futures came out,a large number of financial derivatives emerged,and the demand for quantitative finance was driven by the increasingly complex financial market.Quantitative investment began to attract widespread attention from Chinese investors.The industry and academia were increasingly enthusiastic about it.In recent years,many securities firms in China have set up professional financial engineering teams to conduct in-depth research on the application of machine learning in quantitative investment.In developed capital markets,machine learning has been widely used in quantitative investment.Compared with traditional structured data analysis methods,machine learning have better prediction capabilities for solving noisy and complex financial data.In this paper,random forest,which is a machine learning algorithm that uses Bagging to integrate decision trees,is selected.Taking the CSI 500 index stocks as the stock pool,a total of 39 daily frequency factors are selected from multiple dimensions such as quality,growth,valuation,and risk.To construct a multi-factor stock selection strategy based on the random forest model.The construction of the stock selection model can be divided into two parts.First,the random forest algorithm is used to predict the return rate of the stock on the next trading day;then,based on the predicted return rate,a portfolio is constructed through industry neutralization for backtesting.Based on the initial return rate prediction model,and considering the characteristics of the random forest algorithm,this paper went through two rounds of optimization of feature factor screening and grid parameter optimization to improve the accuracy of prediction.In the end,due to the poor overall market performance during the backtest period,the strategy portfolio only achieved an annualized rate of return of 3.83% in terms of revenue,but the excess return of the comparative index reached 17.25%;in terms of risk,the strategy portfolio had a return volatility The performance of the largest retracement is better than the performance of the CSI 500 index during the same period.