Quantitative Stock Selection Model Based On Machine Learning

With the rapid development of computer technology,wireless communication and the Internet,securities trading has now developed into an electronic bidding and matching method.In this context,quantitative investment is gradually emerging around the world.Multi-factor models are currently the mainstream method for quantitative investment strategies.The quantitative stock selection model of the four methods of regression and support vector machine has achieved good results in the empirical analysis.This thesis selects the daily frequency market data of the constituent stocks of the Shanghai Stock Exchange 50 Index for seven years from 2016 to 2022.First,the data is preprocessed,and then the factors with better stock selection effects are screened out through the factor validity test,and added to the multi-factor selection.In the stock model,and then base on four machine learning methods of CART decision tree,random forest,XGBoost and support vector machine regression,a regression stock selection model is constructed,and grid parameters are adjusted based on the sliding cross-validation method,and an out-of-sample test is performed on the 2022 data.Using the equal-weighted portfolio of the top ten stocks selected by the model,they withstood large fluctuations in the market during the backtesting period,all of which exceeded the index’s return rate.Among them,the XGBoost model has the best test and stock selection effects.The annualized rate of return,maximum drawdown and Sharpe ratio of the combination are 12.1%,25.4%,and 1.32 respectively.The rate of return of the random forest combination is 3.7%,but the maximum drawdown is only 18.9%.Although the combination of CART decision tree and support vector machine regression have a decline of 1.7% and 11.4%,the decline is smaller than that of the index.Although the annual returns of the four investment portfolios have a certain gap,the trend generally follows the market,and the excess returns mainly come from the rising time period,while the trend is close in the falling market,and the maximum retracement is not much different,which reflects the market factors.The huge impact shows that it is not easy to obtain income in a falling market,and it is worthy of further attention and research by investors.The study in this thesis has good guiding significance for actual investment.Combining the multi-factor stock selection model with machine learning algorithms can enhance the interpretability of the model,give investors more new research ideas and methods,help investors increase their wealth,and promote the healthy development of my country’s securities market,maintaining market stability and other aspects are of great significance.