Research On Value Quantitative Stock Selection Strategy Based On XGboost

Traditional value investment uses value measurement indicators and other unquantifiable factors to select stocks through human subjective judgment,while quantitative investment uses models to analyze various factors and select stocks through computer programs.The above two investments have their own advantages.In reality,the factors that affect stock prices are high-dimensional and non-linear.From the past practice,the XGboost algorithm has advantages in dealing with such problems.Therefore,this article attempts to introduce the XGboost algorithm to traditional value investment,construct a value quantification strategy,and study the applicability of value investment combined with machine learning in my country.Based on the analysis of the company’s intrinsic value,this paper selects factors that reflect the company’s value from multiple dimensions and uses the XGboost training model as input variables to construct trading strategies and optimization strategies.First of all,in terms of factor selection,this article conducts an in-depth analysis of the company’s value,and establishes 21 value factors that reflect the company’s valuation,profitability,operating capability,growth capability,and security.Then,seven are effective through correlation screening..Second,in the model construction,XGboost is used to predict the next-period return of the Shanghai and Shenzhen 300 constituent stocks,and the highest 30% is used as a strong stock to build a stock pool.Finally,in terms of strategy,this article uses the top 10% of the strong stocks as value stocks.Through the analysis of the portfolio and the attenuation effect of factor correlation,the position management and holding cycle are optimized,and a good value quantification strategy is obtained.The study found that: First,the traditional value measurement indicators,such as price-earnings ratio,price-to-book ratio and other factors are still valid in the Chinese market,and profitability indicators such as return on net assets and net profit ratio are also effective.Second,the XGboost algorithm has a very good effect on non-linear classification.The introduction of machine learning algorithms in the field of traditional value investment shows that empirical results show that the two-year strategy retracement is only 13%,and a stable value premium can be obtained.Third,value investment in my country’s stock market is effective,and long-term investment is the king of investment by studying the company’s ntrinsic value.The use of machine learning for value investment empirical shows that quantitative value is effective,and the combined research of the two provides a broader perspective for traditional value investment.